Climate models

Tricking yourself into cherry picking.

605 Comments

Did you know you can cherry pick without knowing it? It works like this:

  1. You speculate there “some trees” are temperature proxies, but “other trees aren’t. (So far, you’re actually ok.)
  2. Then, instead of trying to do a real calibration study to discover what sorts of trees are temperature proxies and which aren’t, you just take a bunch of cores and find which correlate “best” with the recent temperature record. You throw away all the rest of the cores as “not temperature proxies”.

This sort of sounds like it makes sense, right? After all, the trees that did not correlate with the current temperature record can’t be temperature proxies. So, the rest may be a little noise, but they “must” be temperature proxies, right?

Oddly, it appears that in comments, this notion
In comments, Layman Lurker shed some light into part of “the argument” between the Lorax and Hu M. Here’s what Layman relates:

IMO Lorax was a bit different. In one situation he asked something to the effect of why it would not be good practice to reject proxies which did not correlate with the insturmental record. Hu M. responded saying (paraphrasing from memory) that “the method you advocate on the surface this seems attractive” but that it would amount to cherry picking and would bias the analysis. Lorax took great offense and went off on Hu in a couple of follow up comments saying that Hu had stated that he (Lorax) advocated cherry picking. I don’t think anyone with any common sense would have interpreted Hu’s response to Lorax’s question that way.

Well… Hu is right. The method of simply rejecting the trees that fail to correlate does automatically bias a sample. Seems odd, but it’s true. So, even though the method seems reasonable, and the person doing it doesn’t intend to cherry pick, if they don’t do some very sophisticated things, rejecting trees that don’t correlate with the recent record biases an analysis. It encourages spurious results, and in the context of the whole “hockey stick” controversy, effectively imposes hockey sticks on the results.

I’m going to show a little back of the envelope analysis that highlights the point Hu was making.

Method of creating hockey stick reconstructions out of nothing

To create “hockey stick” reconstructions out of nothing, I’m going to do this:

  1. Generate roughly 148 years worth of monthly “tree – ring” data using rand() in EXCEL. This corresponds to 1850-1998. I will impose autocorrelation with r=0.995. I’ll repeat this 154 times. (This number is chosen arbitrarily.) On the one hand, we know these functions don’t correlate with Hadley because they are synthetically generated. However, we are going to pretend we believe “some” are sensitive to temperature and see what sort of reconstruction we get.
  2. To screen out the series that prove themselves insensitive to temperature, I’ll compute the autocorrelation, R, between Hadley monthly temperature data and the tree-ring data for each of the 154 series. To show this problem with this method, I will compute the correlation only over the years from 1960-1998. Then, I’ll keep all series that have autocorrelations with absolute values greater than 1.2* the standard deviation of the 154 autocorrelations R. I’ll assume the other randomly generated monthly series are “not sensitive” to temperature and ignore them. (Note: The series with negative values of R are the equivalent of “upside down” proxies. )
  3. I’ll create a proxy by simply averaging over the “proxies” that passed the test just described. I’ll rebaseline so the average temperature and trends for the proxie and Hadley match between 1960-1998.
  4. I’ll plot the average from the proxies and compare it to Hadley

The comparison from one (typical) case is shown below. The blue curve indicates is the “proxy reconstruction”; the yellow is the Hadley data (all data are 12-month smoothed.)

Figure 1: "Typical" hockey stick from generated by screening synthetic red noise.

Notice that after 1960, the blue curve based on the average of “noise” that passed the test mimics the yellow observations. It looks good because I screened out all the noise that was “not sensitive to temperature”. (In reality, none is sensitive to temperature. I just picked the series that didn’t happen to fail. )

Because the “proxies” really are not sensitive to temperature, you will notice there is no correspondence between the blue “proxy reconstruction” and the yellow Hadley data prior to 1960. I could do this exercise a bajillion times and I’ll always get the same result. After 1960, there are always some “proxies” that by random chance correlate well with Hadley. If I throw away the other “proxies” and average over the “sensitive” ones, the series looks like Hadley after 1960. But before 1960? No dice.

Also notice that when I do this, the “blue proxie reconstruction” prior to 1960 is quite smooth. In fact, because the proxies are not sensitive, the past history prior to the “calibration” period looks unchanging. If the current period has an uptick, applying this method to red noise will make the current uptick look “unprecedented”. (The same would happen if the current period had a down turn, except we’d have unprecedented cooling. )

The red curve

Are you wondering what the red curve is? Well, after screening once, I screened again. This time, I looked at all the proxies making up the “blue” curve, and checked whether they correlated with Hadely during the periods from 1900-1960. If they did not, I threw them away. Then I averaged to get the red line. (I did not rebaseline again.)

The purpose of the second step is to “confirm” the temperature dependence.

Having done this, I get a curve that sort of looks sort of like Hadley from 1900-1960. That is: the wiggles sort of match. The “red proxy reconstructions” looks very much like Hadley after 1960: both the “wiggles” and the “absolute values” match. It’s also “noisier” than the blue curve–that’s because it contains fewer “proxies”.

But notice that aprior to 1900, the wiggles in the red proxy and the yellow hadley data don’t match. (Also, the red proxie wants to “revert to the mean.”)

Can I do this again? Sure. Here are the two plots created on the next two “refreshes” of the EXCEL spreadsheet:

Hockey2

Hockey3

I can keep doing this over and over. Some “reconstructions” look better; some look worse. But these don’t look too shabby when you consider that none of the “proxies” are sensitive to temperature at all. This is what you get if you screen red noise.

Naturally, if you use real proxies and that contain some signal, you should do better than this. But knowing you can get this close with nothing but noise should make you suspect that screening out based on a known temperature record can bias your answers to:

  1. Make a “proxy reconstruction” based on nothing but noise match the thermometer record and
  2. Make the historical temperature variations looks flat and unvarying.

So, Hu is correct. If you screen out “bad” proxies based on a match to the current temperature record, you bias your answer. Given the appearance of the thermometer record during the 20th century, you bias toward hockey sticks!

Does this mean it’s impossible to make a reliable reconstruction? No. It means you need to think very carefully about how you select your proxies. Just screening to match the current record is not an appropriate method.

Update

I modified the script to show the number of proxies in the “blue” and “red” reconstructions. Here is one case, the second will be uploaded in a ‘sec.

Hockey4

Hockey

605 thoughts on “Tricking yourself into cherry picking.”

  1. But what if think really sophisticated thoughts before you inadvertently cherry-pick? Doesn’t that cancel out the bias and instead inject penetrating insight?

  2. George–
    I thought of very sophisticated things before cherry picking. I thought:

    Is the correct scientific factor of 1.2, or 1? (Both result in nice cherry picked curves. If I picked 2 I would get even better agreement, but I would need to create many more synthetic”proxies” because I would throw more of them out.)

    I also agonized over deciding whether to permit myself to use “upside down” proxies? Or just “right side up” ones. ( Using the “upside down” proxies gave me twice as many proxies to average in my reconstruction. Since EXCEL is a slug, I used upside down proxies. That’s sort of analogous to using upside down tree-ring proxies because there are so few. . ..)

    Let me assure you I pondered deeply over these choices.

  3. This is great Lucia. So the next issue is how, indeed, to choose proxies correctly. Here are a few thoughts:

    1. Use all available trees and hope that errors cancel.The cherry picked ones will still be used but have a dampened influence (reminds me of an old statistics joke: if you don’t believe in sampling then the next time you go for a blood test ask the doctor to drain all of your blood).

    2. Sample proxies at random and hope errors cancel. Some of the cherry trees will be include also.

    3. Try to figure out why some trees agree with recent temperature and some do not. Base your analysis on the outcome (if there is a solution as to why)

    4. Consider the possibility that tree rings just do not serve as a robust proxy due to factors other than temperature influencing rings. I am sure that the dendro researchers have done a great deal of work in this but it does not seem to be highlighted in re-construction papers a great deal.

    Totally off topic: I recently waded through the documentation for the NCAR Community Atmosphere Model. It is 226 pages! I spotted a few items which indicate that maybe they did not acquire the help of numerical analysts, but I may be wrong (e.g. using Lagrange polnomials for interpolation). My feeling is that anything this complicatd (by virtue of the complicated physical process) may be fraught with hidden errors and ill-possedness issues. Let alone roundoff errors etc. Your opinion would be welcomed.

    Link:

    http://www.ccsm.ucar.edu/models/atm-cam/docs/description/description.pdf

  4. Jack–
    Round of errors actually don’t concern me with codes of this sort. They matter a little, but less than you might think. The interpolation also doesn’t bother me quite so much.

    Ill-possedness is always a risk if it hasn’t been thought out. (But I don’t know if it’s a real problem. I don’t quite agree with Gerry Browning on his notion because… well.. not getting into that.) But the reason I don’t agree with Gerry Browning relates to the thing I do worry about– which is the subgrid parameterizations. I’m not sure that can be done right.

  5. Mann did this form of selection and then used some sort of Monte Carlo technique to show that his reconstruction could not have arisen by chance . Any thoughts on this?

  6. Tag–
    How did he do the monte carlo? (That is: do you have details beyond the word “monte carlo”?

    I’m pretty darn sure that if I supplied you all 154 “proxies” I screened, and you did an analysis, you’d conclude my multi-proxy reconstruction could not arise by chance. Of course it didn’t happen by chance. I screened!!!

  8. Jack–
    The description of the monte carlo is … what word… terse? Cursory? I don’t know if he just screened and then after screening showed the screened results could not happen by random chance (which would mean…. screening is not random), or if he did some sort of monte carlo on a larger population, screened the synthetic results using his method and then showed even then his results couldn’t happen by random chance.

    If he did the first, then what he did achieve a level of bogacity beyond measure. If he did the second, then that’s ok.

    Unfortunately, the reviewers did not think to ask.

  9. Nice lucia.

    Over on CA I reference a paper that selects proxies if the correlation is .4.. or something like that. I think its in the RCS guru thread.

    Also, If you want to do a very interesting study try this.

    1. You now that hadcru global from 1850 to present is made up
    of the averaging of thousands of stations or at a higher level a bunch of 5×5 grids.

    2. Take all the grids. using a 1940 to 2009 window select those grids whose times series correlates with the global series at your threshold. Then average those grids for the time period 1850 to 1940.

    3. Compare the full global index with that projected using the grids that meet your criteria.

    you could repeat this using only the grids that proxy studies use
    rather than the “best grids”

    Now a thermometer is a way better proxy for temperature than a tree. I wonder how well you can predict the past temperature
    ( 1850-1930) from your network of real thermometers.

    I wonder what your error bar would be?

    Just wondering

  12. How many ‘proxies’ were used for the blue and the red curves?

    I suppose the choice of r=0.995 is important – is it realistic (similar to real proxies)?

    The noise in the real proxies – tree rings, lake sediments, etc – may be affected by human activity, which has a pronounced trend in the last 200 years. I think this should make the noise redder in that period, and increase the chances of a hockey stick in the noise (like that Finnish lake thing discussed on Climate Audit). Is that correct?

  13. Interesting post, Lucia.

    Mosher (comment 21647), so even if trees were *perfect* thermometers, you might not be able to estimate a global average surface temperature without a large uncertainty anyway? Nice idea for a study, I think.
    And we know that trees are far from perfect thermometers.
    I learned at CA that some of the trees used in reconstructions grow in climates where temperatures only allow them to grow for 1/10 of the year. A temperature record with missing values for 90% of year. Hmm.

  14. Kusigrosz–
    I don’t know how many. The the method usually threw away about 2/3 for the blue line and more for the red. I can tweak the Excel spread sheet to show that on graphs, but the next three will be new. I can’t get “rand()” to repeat the same thing twice.

    (Note: after writing this answer, I modified the spread sheet and added two new cases. These show the number of proxies in each case.)

  15. Kuzigorz

    I suppose the choice of r=0.995 is important – is it realistic (similar to real proxies)?

    The choice of r matters. The higher it is, the easier it is to get the effect I show. Other things that matter: How many proxies you have to select from, the cut-off you use for throwing away individual proxies etc.

    I don’t know what the value of r is for ‘real’ proxies in any particular study. You’ll notice I don’t refer to Mann in particular. My comment is just illustrating the notion Hu was explaining to The Loraz. How big the problem might be in an application is something that needs to be explored.

    So, the general notion is: When screening you have to be very, very careful. You should check– and the checks should make real sense. This means they need to be explained at some level of detail. At the same time, the fact that there is a general issue doesn’t mean any particular paper is flawed. (I personally don’t read the various hockey stick papers much. But I can comment on some fairly low level arguments, and it’s obvious this Lorax fellow didn’t understand some fairly elemenatary statistical issues.)

  16. I’ve been in market research for most of my adult life, and getting the sample right is the biggest challenge–always. Usually you don’t really want a random sample. What you want is a defined group that covers the characteristics you want to measure, and a selection from that which gives any individual within that selection an equal chance at being included. I cannot image a statistical study where so little is known about behaviour before a certain point (in this case response to temperatures) that would not contain a universe of about 30,000 and a selection for examination of about 1,000. I just don’t see how you do it.

  18. Dr. Rhine in his ESP research did something similar (sheep and goats). When someone scored well they were kept in the experiment (obviously they had ESP), those that didn’t were dropped. Just statisical “runs of luck” made it appear that he had found individuals that had “ESP”.

    Unless you could discern why (what the causation was) one tree matched the temps while others didn’t you’re just doing what Rhine did.

  19. Tom–
    Yes. In marketing research, you might want to find out if “teens in NY ” like “red suspenders”. So, you screen for “teens in NY” and then test to see if that group likes “red suspenders”. What you don’t do is screen by “likes suspenders” as step 1, and then do all your other screens and test second, and then in the end decree that “NY teens do like red suspenders”.

    For tree-nomenters, what they want is “trees whose ring width increase more during high temperature years and decrease less during low temperature years and don’t respond too much to anything other than temperature”. Unfortunately, those trees need to plant themselves, survive a long time and remain more or less equally sensitive to temperature during their entire lifetime.

    Now, if all trees ring widths everywhere in the world responded mostly to temperature and not something else, this would be a slam dunk. You just find any old tree, take a core. do this many places in the world, average and voila!

    But tree ring widths respond to other things: precipitation, disease, tree age, CO2 fertilization, Nitrogen in the soil etc. So, how do you figure out that a particular tree or a group of trees ring widths mostly responded to temperature while avoiding the problems associated with “screening” against the temperature record? It’s not simple.

    It is interesting that some people don’t recognize the general concept that screening the proxies against temperature in the thermometer record introduces certain “features” in the reconstructions.

  20. “This is great Lucia. So the next issue is how, indeed, to choose proxies correctly.”

    There is a good amount of literature on guidelines for selecting proxies that are temperature limited.

  21. Boris–
    Of course. But can you tell us how to select proxies properly for any particular study? And was it done correctly?

    People like The Lorax post things like this:

    Hu McCulloch, if a particular tree ring series is analyzed and it is not consistent with the observed regional SAT record, should it not then be removed from the analysis b/c there is obviously something wrong with it? Why one one be so careless to include a proxy that is not capable of reproducing the observed data record in that region?

    Hu answers stuff like this

    The strategy you advocate is superficially attractive, but unfortunately amounts to outright cherry picking. To see how it can generate the appearance of a dramatic hockey stick using serially correlated artificial data with no true relationship to temperature, read David Stockwell, “Reconstruction of past climate using series with red noise” in Australian Institute of Geoscientists News March 2006 (#83), p. 14, at http://landshape.org/enm/wp-content/uploads/2006/06/AIGNews_Mar06%2014.pdf.

    The Lorax answers like this

    Believe it or not, I am also here to learn. So I take exception to you suggesting that I endorse cherry picking. I was simply asking questions, about how one should go about determining which data are good and bad. I think that we can agree that only good data should be used.

    What is happening is that The Lorax does not understand that cherry picking can happen unintentionally when people don’t understand the effect of improper screening. To devise a method that avoids the bias, you need to understand the bias can arise even though you didn’t intend it.

  22. “But can you tell us how to select proxies properly for any particular study?”

    I don’t know what you mean.

    “And was it done correctly?”

    Is there any way to tell that field work was done correctly?

  23. Lucia: Below the last post(above Tip Jar) are 2 sites for purchasing cherry trees. Is this cherry picking or a coincidence?
    I have held off for a while but I just have to ask: This is becoming a tree-ring circus. Sorry

  24. Boris–
    You said this:

    There is a good amount of literature on guidelines for selecting proxies that are temperature limited.

    and your comment seemed to be responding to someone asking a question about how to pick the proxies. So… can you tell us how to pick them? or what the literatures says? Or tell us whether or not any particular study was done correctly?

    Or are you just posting some sort of apple pie statement? Yes. We all know that “literature exists”. So?

    Jack–Adssense thinks this post is about cherry trees….

  25. Lucia, in addition to Jeff Id, this phenomenon has now been more or less independently reported by Lubos, David Stockwell and myself. David published an article on the phenomenon in AIG News, online at http://landshape.org/enm/wp-content/uploads/2006/06/AIGNews_Mar06%2014.pdf . We cited this paper in our PNAS comment (as one of our 5 citations.) I don’t have a link for Lubos on it, but he wrote about it.

    I mention this phenomenon in a post prior to the starting of Climate Audit, that was carried forward from my old website from Dec 2004 http://www.climateaudit.org/?p=9, where I remarked on this phenomenon in connection with Jacoby and D’Arrigo picking the 10 most “temperature-sensitive” out of 35 that they sampled as follows:

    If you look at the original 1989 paper, you will see that Jacoby “cherry-picked” the 10 “most temperature-sensitive” sites from 36 studied. I’ve done simulations to emulate cherry-picking from persistent red noise and consistently get hockey stick shaped series, with the Jacoby northern treeline reconstruction being indistinguishable from simulated hockey sticks. The other 26 sites have not been archived. I’ve written to Climatic Change to get them to intervene in getting the data. Jacoby has refused to provide the data. He says that his research is “mission-oriented” and, as an ex-marine, he is only interested in a “few good” series.

  26. lucia,

    Is there evidence that proxies in this particular instance were picked incorrectly? I have not been following this thing all that closely, but it seems from my limited perspective that the argument is that if you pick different trees you get different results. This does not seem to me to be a surprising revelation. Do I misunderstand things?

  27. Oh, okay. The “particular study” confused me. IIRC trees are chosen from near treeline. They also select trees that grow farther from other trees rather than trees close together. Obviously they avoid trees with signs of disease, damage, etc. I’m sure there’s other things I’m leaving out.

    I imagine that the paper or chronology would give you information on how the trees samples were collected, but I’ve never been interested enough to look.

  28. Steve
    I should have googled to find everyone who has discussed this. I knew my contribution was not the first!

    The fact is, this concept is so simple, and easy to show. It’s amazing that any researcher anywhere is unaware of the bias that is introduced by screening noisy data.(Or even if they are unaware, that they don’t understand the danger rather quickly.)

  29. Minor point

    “(So far, your actually ok.)” In point 1 of “it works like this”

    should be (So far, you’re actually ok.)

  31. David

    Is there evidence that proxies in this particular instance were picked incorrectly?

    I’m not discussing any particular instance. The topic came up, and the Lorax suggested screening just made sense. In fact, it’s risky and can introduce bias. To know whether it might have in a particular instance, you need to a) name the instance, and b) read that paper and figure out all the screens that happened collectively.

    Whether or not screening caused problems in any particular paper… I don’t know.

    I have not been following this thing all that closely, but it seems from my limited perspective that the argument is that if you pick different trees you get different results.

    Hmm… there are two problems in there. But no, that’s not the argument. The problem is that screening trees whose annual ring growth rates to match the temperature record does not necessarily result in trees whose ring growth actually respond to temperature.

    In the example, the “proxies” were uncorrelated with temperature, but screening still results in a average that looks strongly correlated. That’s because the screening forces the average to be highly correlated.

    This is not to say that screening will never help. It’s just that you have to be very, very careful. The number of ways one can screw up is vast.

  32. David–
    Studies that don’t screen this way are not subject to this problem. (They may have other problems, but not this one.)

    Mind you, you do need to be sure that an author didn’t work from a batch that was pre-screened. For example, if I do a study where I screen and decree a set of trees out of 10 sets to be “a temperature proxy”. Then later you use the set I screened, your results were affected by screening. So, that can be an issue. (And it can complicate how you run any monte-carlo to verify that your results are statistically significant, yada, yada, yada.)

    But if I screened some other way that was not based on correlation with the temperature record, then your proxy would not suffer the problem described in this pot.

    So, connecting this particular issue to any particular paper involves some digging. The Lorax and Hu were discussing some paper that did screen. The Lorax thought it sounded like a rational thing to do. As Hu said, it sounds superficially rational. But really… it introduces bias.

    With regard to Briffa….well, I don’t happen to know. I’m not jumping into all the details about the provenance of all the tree cores. Like Boris ” I’ve never been interested enough to look.”

  34. lucia,

    Fair enough – I simply assumed that because this was about tree rings it had to be about Yamal. 🙂

    On the introduction of bias, isn’t it only the case that it introduces bias if proxies that are accurate today were not accurate yesterday or if proxies that are not accurate today were accurate yesterday?

    As an example, your random series shows that if something matches current temperatures by chance, there are serious problems. But if the match is *not* by chance, don’t the problems disappear? In other words, if the trees match because they are good thermometers, then using them makes a great deal of sense, as good thermometers are what you want to have. So I guess the issue becomes: why do these trees match and those trees not match?

  35. Lucia said “”This is not to say that screening will never help. It’s just that you have to be very, very careful. The number of ways one can screw up is vast.””

    This is so simple and has been lectured in all statistical derived subjects (biology classes, stats class, etc.) that it always boggles my mind when I read it anew. Then to read of how some respond to the criticism of “cheery-picking” when potential problems are shown that were not considered in a paper leaves me management control problems. I think of all those lectures, and really hard and troublesome ways and write-ups I had to do to get a degree and read some of the handwaving … I wonder were my professors sadists to make me do all that stuff the hard way? Or is someone avoiding?

  36. David

    Fair enough – I simply assumed that because this was about tree rings it had to be about Yamal. 🙂

    When people start discussing hockey, all the different studies get interlaced. I can’t tell from the CA thread which one they were discussing.

    On the introduction of bias, isn’t it only the case that it introduces bias if proxies that are accurate today were not accurate yesterday or if proxies that are not accurate today were accurate yesterday?

    It depends what you mean. If they are accidently correlated today, they are unlikely to be correlated during other periods.

    The issue is: Things that happen 5% of the time, do happen 5% of the time.
    Do this: find a chemistry lecture with 1000 students. Have them all flip a coin 4 times in a row. Call the ones who got all heads “skilled coin flippers”; if no on has any skill, there be about 62 of these people. Now make all the “proven unskilled coin flippers” sit down.

    Have the “skilled coin flippers” flip again. Do you expect them to have any particular skill?

    Now, if you start to think about this, you can start to think about ways to figure out whether screening was likely to identify skilled flippers. If you think a bit, you’ll realize that if there are any skilled flippers, there should be more that 62 out of 1000 flippers who pass the first test, right? So, you might devise a test to discover whether there might be any skilled flippers before bothering to screen. Maybe you’d say: If I get 100 skilled flippers, then I’ll screen and repeat.

    So, while this post shows some problems with screening, someone might be able to come up with a way to deal with them. (I bet the Bayesian’s have.)

  37. Lucia,

    thank youfor another excellent post.

    I notice that your FAUX proxies do not reach the magnitude on the HS at the end that the “temp measurement” does. I have also read recently that the dendrochronologists are “interested” in the fact that they have been unable to find a series that matches the magnitude of modern temps.

    Anyone have an opinion on whether RCS, which appears to me to work similarly to what Lucia did, could be part of the issue?? That is, it is selecting non-proxy proxies out of what amounts to noise??

  38. OOOOPPSS!!

    Almost forgot the other similarity.

    The handle appears flattened also. This, according to SteveMcI, in his Briffa writeup, is due to random data averaging out.

    That is, the cores are not actual temperature proxies so they do not move in concert. A large amount of random noise therefore averages fairly flat making a handle.

    With a procedure that selects for correlation during only an end period, if the cores are not actually thermometers, outside the validation period would be more probable not to match and to give a flattened reconstruction also.

    VIOLA

    A complete hockeystick!!!

    Again, any opinions on whether a lot of what we are seeing in Dendro since about 1990 is due to these artifacts???

  39. Good eye Kuhnkat!

    Yes, I noticed my average of faux-proxies don’t peak as strongly in 1998 as the actual temperatures. I could probably capture even that peak if I created waayyyyy more proxies and screened more stringently. But using only 154 and using ‘lax’ screening for individual proxies, no I can’t get that “peak”. BTW: The reason for ending the test in 1998 is to see if I could catch the very sharp run up. I wasn’t catching it with data ending in August 2009, but I was afraid that was due to the “stall” in the rate of increase.

    So, clearly there are two lessons: 1) Screening can create “hockey stickes” but 2) It’s not easy to create perfect hockey sticks.

    Of course, the correct response is to be aware of 1 and consider this issue when reporting the level of significance of any “hockey stick”. The issue is not insurmountable– provided you don’t ignore it.

  40. Lucia,

    You said:

    (Or even if they are unaware, that they don’t understand the danger rather quickly.)

    Do you think there is any possibility at all they were unaware? Honestly, this is very difficult for me to comprehend but not physically impossible. These boys and girls are good at math. RegEM, EIV, PCA, CPS and correlation sorting.

    Here’s a quote from Mann that I find interesting which relates to your post:

    Actually, this line of attack is even more disingenuous and/or ill-informed than that. Obviously, if one generates enough red noise surrogate time series (especially when the “redness” is inappropriately inflated, as is often done by the charlatans who make this argument), one can eventually match any target arbitrarily closely.

    I used his own data in the part 1 post above, but this next quote explains why it’s ok to chuck data that doesn’t support your conclusion:

    What this specious line of attack neglects (intentionally, one can safely conclude) is that a screening regression requires independent cross-validation to guard against the selection of false predictors. If a close statistical relationship when training a statistical model arises by chance, as would be the case in such a scenario, then the resulting statistical model will fail when used to make out-of-sample predictions over an independent test period not used in the training of the model. That’s precisely what any serious researchers in this field test for when evaluating the skillfulness of a statistical reconstruction based on any sort of screening regression approach.

    So he’s hiding behind underestimated autocorrelation in Mann08 for the quality of the result. They know very well what they’re doing IMHO.

    Without Mann tAV might just have ended up discussing interesting science or politics. Instead the guy really caught my attention.

    Thanks Steve Mosher, your comment is appreciated it’s one of my least popular posts but when I wrote it I thought it would be one of the best.

  42. JeffId-

    Maybe so. But, the fact is, screening needs to be done carefully. And the “out of sample” check also needs to be done carefully, otherwise, it’s just a second screen and makes things even worse.

  43. Agreed. I enjoyed the post BTW.

    IMO screening has to happen without looking at the data. There is one exception for data which has obvious errors in collection.

    Explanations of why a mollusk sphincter is or is not a temperature proxy have to happen independently of the data itself. It’s impossible for me to accept any correlation screened data.

    Do you use R or matlab at all?

  44. IMO screening has to happen without looking at the data. There is one exception for data which has obvious errors in collection.

    Or known and explainable contamination of the signal through confirmed physical processes.

  45. Jeff-
    No. I don’t use R or matlab. I either
    a) look up the the mathematical formulations and code it or
    b) use excel.

    R runs so slowly and is interprative. Those two features end up causing me to scream at the screen.

  46. R is frustrating for sure but I can’t remember a software package I didn’t scream at on occasion. Matlab is better but not free so Steve’s right that it’s not as useful for blogs.

    I was curious because compared to excel you can do a lot and there are some interesting details in the code on tAV but I needed to change parameters around a bit to get a feel for it. I’d really like to find a method for correcting correlation sorting, several papers rely on it and it seems to me that knowing the autocorrelation, signal you’re sorting for and data noise level it could be possible.

  47. Jeff–
    It’s true that all software packages involve screaming. I’m sure R is great for lots of things. But why does everyone have to use R to do the simplest sort of things? I’ll learn R when it seems easier to use R than do things ways I already know how to do them.

  48. Tom Fuller (Comment#21657)
    October 15th, 2009 at 3:31 pm
    “… Usually you don’t really want a random sample. What you want is a defined group that covers the characteristics you want to measure, and a selection from that which gives any individual within that selection an equal chance at being included…”

    Tom, the only thing that does that is a random sample. When the required sample size exceeds our budget then we concoct short-cuts, call them “stratified random samples” or some such BS and hope to hell we can justify them. You and I both know that one of the biggest differences between a “random sample” and a “representative sample” is the number of respondents who refused to be interviewed: “….. thank you, dear lady… there are only 27 black, married, Jewish females between the ages of 10 and 80 in Cleveland and the other 26 hung up on me…”

  49. Jeff ID:

    IMO screening has to happen without looking at the data

    Or put another way, you need a set of objective screening criteria that don’t allow feedback from the data you are trying to calibrate to. E.g., if you are looking for a proxy to regional temperature, and you know which tree species and locations for those species are idea for that purpose, you screen for those trees and locations.

    What you don’t do is e.g. use correlation to select out which species you want to use as proxy.

    Lucia, are you saying that excel is faster than matlab? I find that hard to believe.

  50. Carrick–
    No. I don’t have matlab.
    For some things EXCEL is faster than a) finding the R script b) learning R and c) running R. (At least for me.) One of the reasons is the things are simple to do and the other is I already know how to do them. So, I’m not particularly motivated to learn R just to figure out how to do them.

    That’s not to say that I don’t see the value in R. But why learn it just to fit a least squares? Or apply a red noise correction?

  51. Lucia: “No.”

    Whew.

    I was afraid I was going to have to learn to Excel programming now. 😛

  52. I don’t really understand what you’re trying to show here, but I wondered if you clarify your meaning here:

    ” Then, instead of trying to do a real calibration study to discover what sorts of trees are temperature proxies and which aren’t, you just take a bunch of cores and find which correlate “best” with the recent temperature record. You throw away all the rest of the cores as “not temperature proxies”.

    This sort of sounds like it makes sense, right? After all, the trees that did not correlate with the current temperature record can’t be temperature proxies. So, the rest may be a little noise, but they “must” be temperature proxies, right?”

    So the assumption is the trees you are using have not been checked using a calibration study (whatever that is – do you know what sort of calibration study you’d use?). Then you simply check the correlation with the temp record.

    So are you suggesting that trees that don’t correlate COULD be a temp proxy?

    Also, why are you generating random data? Are you suggesting that tree growth is random?

    Or Is this just a hypothetical to show that you can make Hockey Sticks from random noise – and doesn’t actually have anything to do with trees at all?

  53. Is there a theory in climate science anywhere that predicts accurately what a tree core will show based on the existing temperature record? That is to be able to say, the core will look like this after I collect it and then it does.

  54. Also I think you need to define what you mean by Cherrypicking, becuase this statement:

    “The method of simply rejecting the trees that fail to correlate does automatically bias a sample.”

    Is, well, not that interesting. Whenever people deal with natural data they will make selections based on criteria. So it’s important here to define what you mean by Cherrypicking as that has an implication that you are selecting data base on a desired outcome. Sometimes the bias in selection is what you want! For example when people do geochronology they preferentially select samples, so yes there is a selection bias, but no one would suggest that geochron is cherrypicked to give a date for a rock.

  55. I am kind of with Nathan here. The random series does not seem to relate to tree proxies. It is not all that likely that a particular tree will correspond to the recent temperature record by chance, given that there are many possible ring growth outcomes and that ring growth outcomes are not random – they are caused by, among other things, temperature.

    So: if you find trees that correlate well with the recent temperature record, it would seem to me that the most likely reason for that is not chance but because those trees are indeed good proxies for temperature (at least within the recent time period: you cannot *know* whether or not they are good proxies for the time period that you are investigating, as the whole point is to use these tree rings to reconstruct something that you do not know – although of course if the proxies show something outrageous, like the temperature was 100 degrees centigrade, then you have some clue that there is a problem … ;))

    If, of course, it could be shown that trees showing a poor match with known recent temperatures should be included – perhaps because whatever caused the poor match was not in effect during the period that you want to learn about – then they should be included. Likewise, if it could be shown that trees showing a good match for recent temperature should be excluded because they would have been affected by something other than temperature in the period that you want to learn about, then that will help give a better picture.

    If the random sampling idea is that the bad proxies will cancel each other out and leave a reasonable approximation, the problem is that if there are 50 bad proxies and 50 good proxies, the signal will be significantly dampened by the noise, even if the noise is averaged. It is much better to search for good proxies and eliminate bad ones if at all possible.

  56. Lucia said:

    Yes. In marketing research, you might want to find out if “teens in NY ” like “red suspenders”. So, you screen for “teens in NY” and then test to see if that group likes “red suspenders”. What you don’t do is screen by “likes suspenders” as step 1, and then do all your other screens and test second, and then in the end decree that “NY teens do like red suspenders”.

    You know, I started digging around on climate blogs a few years ago, not because I find the science fascinating, but because I’ve been alive long enough to know that when someone says “The end is near…” quickly followed by “Send me your money”, it’s probably time to turn up the BS detector gain and start scratching around for some data. I’ve been following these proxy reconstruction discussions across the spectrum of climate change blogs. I was tormented by enough required stat while making my way through a math/physics curriculum to follow things reasonably well, but I’ve always asked myself how one could translate this arcane pabulum into something that the un-tormented could relate to. Lucia has delivered the answer in the brief sentence quoted above. Most anyone that wants to understand what Briffa did with Yamal data doesn’t need to look any further. Or perhaps he just used the Russians to fight a proxy war? Da-da-bing 🙂

    OA

  57. Jeff Id (Comment#21687) October 15th, 2009 at 7:16 pm

    Jeff I thought the post was just brilliant. from one old engineer who’s had several lobotymies ( the first required for marketing, the second for managment, the third for beoming an executive) to another who has obviously retained all his marbles.

  58. On another side issue, I am not convinced this constitutes a Hockey Stick by the definition you started to give on that earlier thread.

    There was a requirement that the data be a proxy for temperature. Random data isn’t a temp proxy so this isn’t a Hockey Stick.

  59. You have made basic flaw in your logic. It’s like saying because Michael Mann owns a gun, and guns can be used to kill people, Michael Mann has killed people with his gun.

  60. Nathan and David I don’t think Lucia tried to claim that her random data was a temp proxy but only to show the issue of selction bias (and that is bias in the stats sense not political sense). Basically you have to be (as Lucia has said) very careful if you are throwing out data (David as we have seen in previous threads here), if you just use the a recent correlation as your filter you leave a fair chnace of an error especially if your recent period is small compared to the overall study lenght (say the last 100 years compared to a 1000 years). The random runs just show that a hockey stick shape is possible even in the absence of causuality, so a good by eye correlation to a hockey stick shape is an insufficient test.

    It seems to me with much of the climate blogwar the Stats people often find ground for critising methodology especially when published work is insufficiently explicit about the data, its provenance or the analysis methods. Sure there may be right and proper grounds for data selection but it is good if that is made clear and the filter by recent correlation is not sufficient.

    Of course when a stats-head like Lucia says a method is biased I don’t think she is making no moral judgement only a stats judgement.

  61. See comment on not using the correlation between data and proxy as a feedback to improve your proxy.

    Doing so is an error, period. Even if your last name is Mann.

  63. Andrew

    My post about this not being a Hockey Stick, was tongue in cheek… Forgot the smiley face.

  64. #2713. You quoted a post from March 2008. And yes, you’d think that bristlecones would have disappeared from the conversation long ago. But their use continues unabated, including most recently Mann et al 2008. Indeed, the use of strip bark and even Mann’s PC1 actually continued after the NAS panel recommended that they be avoided in reconstructions: for example, Hegerl et al 2007; Osborn and Briffa 2006; Mann et al 2007; Wahl and Ammann 2007; Mann et al 2008. Yes, the issue should have been put to bed a long time ago. However, the people that you should be asking about this are the people at realclimate, not Lucia and myself.

  65. Nathan, David, IMO there is nothing wrong with following a pre-defined protocol to select trees for sampling which are good candidates to express the temperature signal. Once the sampling is completed on these trees, you will have a statistical “population” – a frequency distribution of the values sampled. In a straight forward case you might get a well defined normal distribution (the Gaussian “bell curve”) where the peak frequency (mean) corresponds with the temperature signal. Values on either side of the mean (+ and -) could be thought of as signal plus random noise.

    Conversely, you may have a population distribution that does not correspond with the temperature signal (divergence). In this case there may be values which correlate with temperature but the population distribution shows us that it is much more likely to be a “divergent” tree plus noise.

    It is also possible that the sampled trees contain two populations – one corresponding with divergence, and one corresponding with temperature signal. In this case the frequency plot of sampled tree values would have a two peaks associated with each population. If the distributions are overlapping significantly, then there will be both signal and noise correlating with temp.

    The important thing in all of this is that the complete population provides the context with which inferences can be made. IWOW, the level of confidence we have that a population reflects the temperature signal.

  66. Jeff Id said (in re to his blog post http://noconsensus.wordpress.com/2009/06/23/histori-hockey-stick-pt-2/) ” …it’s one of my least popular posts but when I wrote it I thought it would be one of the best.”

    “Least popular” and “best” aren’t necessarily mutually exclusive. Your post is what motivated me to download and install R and start to learn it.

    Your post showed that Manns’ own dataset and algorithm can be used to generate just about any arbitrary result you want to generate. What is incredible is that so many people independently rediscover this, yet in the peer-reviewed literature, hockey sticks live on.

    My general rule is not to assume evil motives when ignorance or sloppiness suffice as an explanation for a behavior. At some point, the ignorance hypothesis is falsified.

  67. Oh yeah, I agree Layman. There is certainly a need for a pre-defined protocol, which it seems there is.

    I am just a bit wary of the word(s) Cherrypicking. That has a colloquial meaning that doesn’t entirely related to what Lucia is posting about here. That’s why I wanted her to define what she meant by cherrypicking.

  68. Lucia,
    As already mentioned, this has been done in various slightly different forms by several other people, for example lubos at
    http://motls.blogspot.com/2008/12/ccnet-eight-articles-about-climate.html

    Here is my matlab implementation of the effect, averaging the ‘proxies’ that have a correlation above some threshold with the ‘true’ temperature.

    templength=100;proxylength=1000;nproxies=200;corrthresh=0.3;
    truetemp=(1:templength); % ‘true’ temp record, linear
    recon=zeros(1,proxylength);
    for n=1:nproxies
    seq=rand(1,proxylength)-0.5; % random data
    for k=1:proxylength
    seqb(k)=sum(seq(1:k)); % make a random walk
    end;
    seqend=seqb(proxylength-templength+1:proxylength);
    c=corr(truetemp’,seqend’); % check correlation of proxy
    if c>corrthresh
    recon=recon+seqb; % include proxy if it passes the test
    end;
    end
    plot(recon) % the magic hokey stick

  69. As an alternative to rejecting the uncorrelated. How about only rejecting those that are uncorrelated at high frequencies. For example: is the year to year difference in temperature correlated with the year to year difference in ring width. I have not tried it, but it would be a simple method that got rid of some uncorrelated trees without imposing any low frequency signal.

  70. Lucia,

    Excellent post, it has helped confirm to me (very much a layman in terms of statistical analysis) the basic principles underlying many of the technically ‘deep’ postings on this and other sites.

    Cheers

    Mark.

  71. Aslak–
    That might work. Obviously, one wants to identify trees whose growth does respond to temperature and exclude those that don’t. However, one must do it in a way that does not, itself, introduce a signal (not even if you think that signal is “right”.)

    We might be able to test “correlation at high frequency” vs. “low frequency”, but I’m not quite sure how to do that this second. (Just woke up”.

    Nathan–
    This is the definition of cherry picking:

    Cherry picking is the act of pointing at individual cases or data that seem to confirm a particular position, while ignoring a significant portion of related cases or data that may contradict that position.

    Screening can do this. It cause it to happen inadvertently making “noise” support the position the researcher holds. Oddly enough, because the researcher did not do it on purpose, he often feels very confident in his result.

    I suspect most cherry picking in science is unintentional. People think their reasons for screening are proper.

  72. Aslak Grinsted (Comment#21720)

    The problems with unintentional cherry picking can arise even if some of the proxies are “signals” and not just red noise.

    In an earlier CA comment (#158), I pointed out that:

    If in fact there are differences in “temperature sensitivity” …, then picking sensitive proxies only for the current time period would in fact raise the possibility of serious bias because the current sensitive proxies are being compared with a mix from population with a wider range of lower sensitivity thereby depressing the response of the sample during the earlier period.

    In such a case, the proxies would give a good account of the temperatures during the instumental era. However, since the earlier proxies can not be screened similarly, the risk of biasing the earlier part of the reconstruction in the same way as Lucia points out still exists.

  73. Lucia

    OK that’s a fair enough definition.

    I don’t see how it relates to the specific instance you quoted:

    “In one situation he asked something to the effect of why it would not be good practice to reject proxies which did not correlate with the insturmental record. Hu M. responded saying (paraphrasing from memory) that “the method you advocate on the surface this seems attractive” but that it would amount to cherry picking and would bias the analysis. ”

    Because obviously part of checking if the tree is a proxy is comparing it to the instrumental record. It seems entirely senible to discount trees that don’t hold up to scrutiny. Remember we can’t tell which trees are proxies, we can only see which ones have the appropriate characteristics. Seeing if they fit the temp record is just part of the testing process.

    Same thing happens in U-Pb geochronology. They crush a rock and extract as many zircons as they can. Zircons are then checked to see if they are suitable (based on clarity, zoning etc). Bad ones are discarded (or bad trees not sampled because they know they are not going to be good thermometers). These remaining Zircons are dated and those dates compared to Stacey and Kramers’ (1975) curve. If they don’t fit the curve it is assumed that some other process has disrupted the crystal and upset the isotopic ratio (Or sampled tree data that doesn’t fit the temperature record). The remaining crystals (that fit all the criteria) are used to make the date.
    If we know that a tree, to act as a proxy, must match the temperature record, why can’t that be an exclusive feature?

  74. I guess the key part of your definition is this statement

    “while ignoring a significant portion of related cases or data that may contradict that position.”

    In the tree-mom-meters case the decision to exclude that data that contradict the position is justified (in the case of data not matching the instrumental data). Why would you include data that you know is bad with the good stuff?

  75. Instead of winnowing the data, the obvious (to me at least) approach would be what an experimentalist would do. The hypothesis is that trees are “treemometers”. Some trees aren’t responding in the expected way. Generate a hypothesis that there is another factor or factors that prevent the trees from responding as “treemometers”. Investigate other parameters that may affect the results. Determine if the trees that don’t respond belong to a distinct population such as location, soil ect.

    If you can’t then you have to reject you’re original hypothesis as unproven. Of course you may not have the data to do this sort of evaluation, but that doesn’t seem to stop anyone.

  76. Nathan–
    Clearly, you don’t understand the problem.

    Zircons are then checked to see if they are suitable (based on clarity, zoning etc). Bad ones are discarded (or bad trees not sampled because they know they are not going to be good thermometers).

    Note this screening is not based on the correlation between what a zircon metric and what you think the date ought to be based on, say, the date on a fossilized clay tablet whose date you wish to confirm and which happens to be buried along with the zircons.

    The screening is based on clarity, zoning etc. These requirements were determined in previous studies using different zircons. This is an important point: Screening should not be based on correlating with the thing you are trying to reconstruct.

    These remaining Zircons are dated and those dates compared to Stacey and Kramers’ (1975) curve. If they don’t fit the curve it is assumed that some other process has disrupted the crystal and upset the isotopic ratio (Or sampled tree data that doesn’t fit the temperature record). The remaining crystals (that fit all the criteria) are used to make the date.

    Assuming you are describing this correctly, note that the the screening is not based on matching the correct date printed on the clay tablet buried with the Zircon’s and whose date you wish to confirm. It’s based on an entirely external reference curve that as been shown to apply to zircons in any setting.

    If we know that a tree, to act as a proxy, must match the temperature record, why can’t that be an exclusive feature?

    Because if what you assume to be proxies are noisy, your population of “proxies” may be dominated by samples that exhibit spurious correlation.

    This is why you need a screening method that is not based on the result you are trying to tease out of the proxies.

    No one says just pick any old thing at random to use as a proxy. The issue is that if you want to use proxies to reconstruct temperature, you should not use correlation with thermometer record to screen the proxies.

    What you might do is establish that certain types of trees always have very good correlation with temperture. Maybe you identify that “Truffle trees” everywhere in the world under any and all circumstances are good proxies in study “A”. Maybe these use truffle trees cores 1-1,000,000 ever sampled. This could be a calibration study. You might find a screeing practice that helps remove noise. (For example, you might find that it works as long as the tree is between 10-400 years old, but less than 20 years prior to death. Or it works provided the trees are riddled with insect holes etc.)

    You create “rules” for the screening future groups of truffle trees to use as proxies; these rules should not include individual trees matching the current temperature record.

    Later to create a historic reconstruct, you use trees that were not used in study A. You sample all available truffle trees, screening according to the rules outlined in study A– which did not screen to match the current temperature record/

    Now, if your new set of trees does match the temperature record, you have out of sample confirmation that the truffle trees “work”, and you have more confidence in your proxies.

    There are other alternatives one can do. But the difficulty is that screening of individual samples to make sure those specific samples give you the “right result” will cherry pick your sample to include only those cases that give the “right result”. Since that method of screening can create the “right result” out of totally random samples, you can’t conclude anything from a result that “looks right”. The result could either be the result of screening or the true result. The problem is there may be no good way to know which is which!

  77. Nathan, do you really not understand the problem? I’ll try to explain it one sentence:
    The proxy screening process is guaranteed to generate a hockey stick picture, whether the input data is tree rings, stock prices, or random numbers, therefore the results of such a study are meaningless.

  78. Bugs,
    .
    People have been mostly ignoring your comments, and with good cause. But in the hope that you can learn to rationally critique a scientific study, I offer you the following comments on Briffa’s Yamal study. Please think about the issues for a bit before replying.
    .
    The potential problem with a proxy study is that inappropriately applied selection criteria can generate false correlations. Lucia has correctly pointed out in her post that even random noise, combined with an inappropriate selection method, will generate a false correlation with a historical record.
    .
    In the case of Briffa 2000 (Yamal data), some have speculated that trees which lived (and some still living) during the instrument temperature record may have been selected because they showed correlation with the instrument temperature, but we do not really know if that is the case. There are no doubt legitimate selection factors (not rotting, no evidence of fire damage, etc.) that have to be applied. The 12 selected trees for the most recent period (the “dirty dozen” as some have called them) were all very old, and grew for a long time before as well as during the instrument temperature record. No younger living trees (100 years old or less for example) were selected to represent the most recent period, and the total number to trees for the recent period is very small (much lower number than for earlier periods) even though living trees are common and samples are easy to collect. This single fact, without a clear explanation and justification, ought to be picked up by anyone’s BS antenna. Briffa DID NOT disclose the very small number of trees representing the recent period. If he had, then most people would automatically doubt the results. Neither the Russians who originally collected the tree samples nor Briffa have offered a clear description of their selection method. But whatever that method was, it excluded all but the very oldest trees for the most recent part of the record.
    .
    Briffa then compared the behavior of these selected very old trees to a calibration growth curve generated from the complete record of tree samples (mostly long dead) representing several thousand years. The problem is that the selection method used to cull the living tree population down to 12 very old trees clearly WAS NOT applied to the complete data set (since the complete data set includes trees of all ages), and the calibration curve included a broad mixture of tree ages. So the calibration curve was generated from a population that is not representative of the selected trees for the most recent period. If the Yamal trees really do respond to temperature increases by increasing growth rates, then younger living trees should also show the same temperature driven pattern as the oldest trees.

    Whatever selection method is used in a proxy study, it must be uniformly applied to all the data to generate a valid correlation, and the selection process certainly should NOT include correlation with the instrument temperature record, since the same selection can’t be applied to samples from before the instrument record. There is no doubt that the 12 very old trees in Briffa’s study do show more rapid growth in the 20 century compared to that expected from the calibration curve, but without a description of the selection process and without the inclusion of a full range of tree ages for the instrument temperature period, it is impossible to say that the 12 very old trees are really showing a temperature trend instead of something else. Maybe they are showing an increased growth from higher CO2 in the atmosphere, or maybe they just outlived their neighbors and finally had “their years in the sun” when the neighbors fell. We just don’t know, and Briffa is not talking.
    .
    A study like Briffa 2000 is terribly weak due to poor selection of data. The problem IS NOT that the 12 selected trees do not correlate with increasing temperature during the last 100 years (we have the instrument temperature record, we don’t need to look at tree rings), it is that it is impossible to conclude what the pre-instrument temperature really record looked like based on Briffa’s results.

  79. PaulM, thanks for the code snippet.

    For people don’t have the statistical toolbox, replace this line:

    c=corr(truetemp’,seqend’); % check correlation of proxy

    with these:

    rr = corrcoef(truetemp,seqend); % check correlation of proxy
    c = rr(1,2);

    (corrcoef is a standard Matlab function).

  80. This simple concept of biased selection has been explained at length in basic statistical texts and papers, numerous blogs have demonstrated the principle with concrete examples, any person with basic statistical training understands the problem. People like Nathan and David simply do not have the statistical education necessary to understand the problem, a characteristic shared by most laymen. I had only a few courses in statistics, but have been shepherding my sons through High School statistics this semester. They have had to learn that what seems to be logical, in light of their ignorance, may not be statistically valid practice. Jeff ID’s article was a valuable teaching tool, as is your demonstration here.

    When I read the comments of otherwise seemingly intelligent persons such as Nathan and David (not to mention climate “scientists” who do not understand), I despair of most laymen ever understanding the issue. I salute you for continuing to try to enlighten the ignorant.

    Regards,

    WJR

  81. Nathan:

    If we know that a tree, to act as a proxy, must match the temperature record, why can’t that be an exclusive feature?

    The short answer is you can’t just cull data this way. It’s an error.

    If you know what your data are supposed to measure (and generally they do), you throw away data points that don’t fit with what you “know” to be the right answer. That’s cherry picking at it’s core regardless of why you choose to cull out part of the original data sample.

    There is an undergraduate laboratory experiment that illustrates this, measuring “big G”, the gravitational constant in Newton’s Law of Gravitation between masses m1 and m2 with seperation r12:

    F = G * m1 * m2/r12^2

    that nicely illustrates this. (This law applies to point particles, and by extension to the exterior problem for spheres.)

    The student uses a suspended cylinder on a torsion balance and measures the angular deviation of the torsion balance as a function of the separation between the cylindrical test mass and a spherical test mass.

    Several students inevitably come in with a number to within three digits to the published value of G at the time.

    Well there’s a problem with that.

    There’s a systematic error that the students weren’t told about: The force of attraction between cylindrical mass and a spherical doesn’t exactly follow Newton’s law, intended as I mentioned for point objects. (There’s also a tilt sensitivity of the torsion balance that couples in, because as you move the large mass closer to the torsion balance, the floor tilt increases, among other systematic effects that makes this the toughest constant to measure accurately.)

    Anyway, after the students have finished their work, the professor explains the problems (if he’s on the ball he gives them the corrected formula and has everybody recompute their constant with that…the answer will still be systematically off due to the other systematic effects that haven’t ben controlled for.)

    There are always a few red faces in the classroom trying to explain why their answer was accurate to 3 or 4 digits, when they shouldn’t have (at least without all of the other corrections).

    I think what Mann is doing, in the best of intentions but regardless, is basically this. Except he’s correlating tree ring data to global land+air mean temperature, and there is no theoretical basis for this, it’s just an assumption.

  82. re: SteveF

    I think you might have touched on a legitimate issue with the Yamal chronology. RCS standardization fits a single curve to the proxy data. The curve is supposed to represent the biological response to climate for the “population” taking the age of the trees into account. Sub-groups must fit the same curve for validity.

    Here is a quote from: Esper, Cook, Krusic, Peters, and Schweingruber, “Tests of the RCS Method for Preserving Low Frequency Variability in Long Tree Ring Chronologies”; Tree Ring Research; Vol. 59(2), 2003

    “RCS is clearly sensitive to the effects of different subsample populations entering into the calculation of single RC. Including samples from different “biological growth” populations in one RCS run could bias the resulting chronologies (e.g. TRW in Figure 8C) thus affecting interpretations of climate made from resulting chronologies. However, opportunities to test the data for existence of different populations are limited. This dilemma originates from the condition that the RCS requires only one RC for all the series then calculates anomalies form this one function for each single series. This approach works like a black box, making latent defects during the standardization process difficult to detect. Such defects can be studied and corrected much more easily when each single series is standardized individually. We recommend separating the data into possible subsamples then analyzing (i) the raw chronologies, (ii) the mean curves after age realignment, and (iii) the relationship of the mean versus the age of individual series. The classification of the population subsamples might follow the meta-information of a collection, and should certainly consider such differences as dead vs. living trees, site ecology and species composition.”

    Jeff Id posted the RC fit for the CRU 12 and compared it to the rest of Yamal: http://noconsensus.wordpress.com/2009/10/05/yamal-the-dirty-dozen/

    This seems so simple that I feel I must be missing something.

  83. Lucia,

    This problem is very much understood in machine learning where a great deal of effort is expended in understanding: for any ‘machine learning’ method what is the probability that the r^2 correlation of out-of-sample data is within some tolerance around the r^2 of the training data.

    One of the very basic results here is that if you try more than a handful of models, your confidence-intervals explode–this is in fact my chief concern toward Jeff Id’s antarctic reconstruction work. He keeps trying to tweak his method to goose the correlations. That sort of thing is just horribly, horribly wrong unless you use different data to develop your method than you ultimately use in the reconstruction.

  84. Correlation is not causation. Without knowing the causation, you can go wildly wrong polishing the correlation.

    The basic problem with temperature proxies is that they are desirable for finding the temperature before thermometers existed. But the mechanism of the relationship should be known for each sample, because what should be calibrated to the temperature record is the mechanism and not the sample.

    If you don’t know why something is correlated to temperatures, you can’t properly compensate for various effects.

  85. WJR (or anyone else),
    Do you know of any textbooks that explain this effect?
    The problem is that some people won’t believe anything they read on skeptic (or even lukewarmer!) blogs.
    I find it is one of those things that can be puzzling at first, but when you understand it, it’s obvious.

  86. Lucia,
    any idea what the exact further steps are that they take to verify a correlation between the tree and modern temperature is indeed temperature driven dependence (cause and effect and not mere coincidence)? One weak test but good reality check would be to figure out how many random samples would correlate based on chance alone and then see if their samples’ correlation exceeds that.

    Of course, even if they have a good method of verifying MODERN temperature dependence, how can they possibly verify that the very same tree was ALWAYS a good thermometer BEFORE the instrumental/verification timeframe? There’s no way to verify. I have to recalibrate test equipment annually for this very reason; you can’t rely on calibration during one period to validate calibration status during another!

    Add on top of this, you also have trees that lived entirely out of the instrumentation periods altogether; how do you pick which samples to use among those? There’s no thermometer data to compare/verify for even a portion of their lives. At best you’d have to compare a portion of those tree’s life that overlapped with your supposed modern thermometer trees. Now you’ve of course compounded error and greatly reduced your confidence in the results.

    And if you’re picking pre-calibration trees differently than the contemporary trees you are changing the mechanics of the “thermometer” itself. What are the effects?

  87. hhhhmmm,
    No, I don’t know what exact steps “they” take. I suspect it varies from paper to paper. That’s why I can’t say that this analysis is related to any particular paper– it merely relates to the discussion bewteen The Lorax and Hu. The discussion was on the general issue. If you screen based on correlation between temperature during the current period, you can (and will) bias results. (Quantifying the bias would be difficult and would would then need to look at individual cases.)

    Hu, a statistician, was correct on this point. The Lorax was wrong.

    One weak test but good reality check would be to figure out how many random samples would correlate based on chance alone and then see if their samples’ correlation exceeds that.

    But, yes. It seems to me that this would constitute a reasonable, but possibly weak test. For example, you could apply it to detecting the number of psychics in a chemistry laboratory of 1000 students. You have all of the student predict the face on 4 coin tosses. By random chance, 1000/16 = 62 should get the correct result. Before you even bother to try to “screen” to find psychics in the group, you should make sure that at least 75 students guessed right. (This will happen in 1 in 20 cases. Thinking two-tailed, its the 90% confidence interval for what happens by chance.)

    So, if you get 75 students, after than you should screen, and then either a) retest the 75 students or b) retest everyone.

    With the trees, I think the best you could do is study loads and loads and loads of “truffle trees” during the modern period and find out whether they act as good tree-nometers everywhere. After all, if you had thousands of truffle trees and every single one calibrated well under all conditions, you would pretty much believe the data from ancient trees even if those for some reason, cored back in 1800 and the cores were stored. (But then, you wouldn’t really have to screen, would you?)

  89. Lucia,

    The screening you describe is the rational in Mann08. 484 of 1209 proxies selected by correlation whereas only … if not. I can’t remember the lower threshold. However, the proxy series were hand selected before use. Data which didn’t conform before correlation was chopped off many of the accepted series (schweingruber MXD). Full record proxies (like 51) were then used to regress the data back on the ends of 100% of the (1209 minus 51) other proxies. Of course the 51 proxies had a strong upslope so he pasted blades on the stick handles.

    Now we who have used correlation know that strong slope= strong correlation, it’s not linear. So then by adding the upslope you can receive a high number of proxies through the correlation filter.

    What’s worse is that 71 of the series were actually temperature data (Luterbacher) which of course all but one passed the correlation test.

    Finally, they used one correction for autocorrelation. I haven’t investigated it fully but we know that autocorrelation is the recognition that serial correlation of data results in improved correlation stats. I’ve tried to calculate AR values for the entire series and it’s all over the place with some series refusing to converge in the R ARIMA function.

    Suffice it to say, there is more than a few problems in Mann08’s version of the quantification of bias.

    My own estimate is Mann08 = 100% bias.

  90. Nathan, how do you know your good trees were good thermometers in the past?

    Plus, in the case of Yamal, these are short lived trees.
    So in the current time frame you have 50 good trees,
    while in the past you have a larger sample of trees, some good, and some bad. Out pops a hockey stick.

    In theory, you could take your good trees, assume they are good in the past if they are good now, then match old trees up against the record of the good trees.

    There still is the problem of how we know this is not random chance. That trees respond to temperature is fine. How much correlation factor are we talking about here? .3-.5.
    Is that good enough?

  91. JeffId–

    One of the reasons I don’t read many of the details about all the Mann stuff is that… well… it’s trees… There is so much noise in that, period.

    Out of curiosity, when you say the Luterbacher series is temperature data, could you elaborate? Are they trees? Or are they, literally thermometers? Or what? I have no idea what that series is.

  92. Can anyone tell me a method of determining exactly if a tree’s growth is solely due to variations in temperature and more importantly, without sampling it.

  93. Lucia, the Luterbacher “proxies” are European gridcell temperature reconstructions, which, as I recall, directly incorporate instrumental temperature in the instrumental period, extending it back using documentary evidence. They, unsurprisingly have enormously high correlations to CRU gridded temperature. If one does a histogram of M08 “proxy” correlations, they are a distinct group way off on the right side of the scale. It is absurd that they should have been included in an assessment of the “proxies” – they obviously make them look better than they really are. And there are quite a few of them in the 484 – between 75 and 100 as I recall.

  94. Lucia, I can’t find my copy of the full paper but here’s the quote from MannSI.

    No instrumental or historical (i.e., Luterbacher et al.) data were used in this procedure.

    European temperature reconstructions of Luterbacher et al.
    (3) available at 0.5° latitude/longitude resolution and at monthly
    (A.D. 1659–1998) or seasonal (A.D. 1500–1658) temporal resolution
    were upscaled to a 5° grid and annually averaged
    (Jan–Dec) to yield 71 distinct annual grid box series back to A.D.
    1500.

    Here’s a link to the paper in question.

    http://cat.inist.fr/?aModele=afficheN&cpsidt=15556748

    Also, my recollection is that proxies were pasted on the end of temp data through math. Beyond that, I’d have to dig up the paper again.

  95. SteveM and I crossposted but we’re saying the same thing. He was the one who pointed it out to me though.

  96. Cherry-picking is more formally called data mining. For a simple description of wht can go wrong in a completely different area try a paper by Grant McQueen and Steven Thorley

    http://marriottschool.byu.edu/emp/Grm/Finishedpapers/AAIIGalley.pdf

    From the conclusion

    “If you torture the data, it will
    confess. Statistically significant
    but spurious patterns occur in
    data simply by chance. Thus,
    one should not be surprised to
    find “word clusters” in the Bible
    after searching over enough
    equidistant letter sequences, or
    cancer “hot spots” after
    searching across enough towns
    and enough types of cancers.
    Even finding that stock returns
    are related to the phases of the
    moon is not surprising, if the
    finding came at the end of a
    long search of unrelated nonfinancial
    variables (see Timothy
    Crack’s, “A Classic Case of Data
    Snooping,” in The Journal of
    Financial Education, Fall 1999).”

    And ealrier they desribe how data- mining can give you good ideas but you must then test again. Smoking and heart disease is a good example.
    “the smoking/heart
    disease case is an example of
    beneficial mining–mining that
    cleared all three hurdles
    discussed above. First,
    researchers readily disclosed
    the number of variables
    searched over before finding the
    link to smoking. Second,
    subsequent work developed a
    clear and concise story or
    theory about the mechanism
    that linked smoking and heart
    disease. And third, the link held
    up time and time again in outof-
    sample tests with new
    subjects not in the original
    studies.”
    Most of the correlations studies on tree rings and the like do not match up to these standards. Sometimes they tell you how many variables they searched over, but often fail to account for the autocorrelation in the data when they work out how many “significant” variables they find. They never establish a rigorous mechanism and most of the claimed relationships don’t stand up outside the original study period – i.e. the divergence problem.

  97. Ok– So it includes *direct temperature* measurements? Yes. Not a proxy. (Possibly useful information, but clearly, not a proxy.)

  98. mikep

    Most of the correlations studies on tree rings and the like do not match up to these standards. Sometimes they tell you how many variables they searched over, but often fail to account for the autocorrelation in the data when they work out how many “significant” variables they find.

    What’s not clear to me (and it may just be my failure to study what everyone did)…. is this: When a researcher looks at data in the archives, does he know how much prescreening was done previously? And precisely how it was screened.

    In my “truffle tree” example above, I mentioned that if the trees 1-1000 were used to demonstrate that “truffle trees” are temperature proxies and that a certain screening method (based on lack of insect infestation etc.) seems to work, you should avoid those in a later reconstruction.

    Instead, you need to use fresh brand new truffle trees, apply the screenings and make the reconstruction. That way, you know that you didn’t just accidentally screen for a good match during the current period. If you got a good correlation after screening, this would confirm the previous result– it would basically just be noticing the already existing screened archive didn’t magically change properties while in storage.

  99. stephen richards #21761
    Yes – Put it on your list of goodies for Santa Claus. But remember, that only works for boys who have been “good”. Otherwise you get coals!

  100. Lucia, in your example of truffle trees, how would it work if you had two factors competing for the truffle tree growth. For instance if fertilizer and temperature availability both increased at the same time.

    I think that the assumption you make requires that all non-temperature influences must be orthogonal. On average at least.

  101. Rethinking this, perhaps anti-correlated ‘other factors’ would work as well, but non-stationary relation yet potentially correlated factors such as CO2 would cause problems.

  102. JeffId–
    I suspect that if two factors affect truffle tree ring widths, the and you did did your various background studies poperly, you would dicover that truffle trees are, quite simple, bad tree-nometers. Sometimes they would respond to temperature; sometime they wouldn’t. That’s the way it works with measurement devices.

    Good devices have something you can measure that responds to the thing you want to know.

  103. Your reaction is the same as my own. Trees make lousy thermometers. It’s very difficult to tease a real signal from noisy data. If you don’t have explanatory mechanisms for the noise, it’s almost hopeless to sort and improve on a simple mean of the entire set.

    The power of mean is difficult to ignore.

  104. Layman Lurker (Comment#21741),
    .
    “This seems so simple that I feel I must be missing something.”
    .
    Nope, you have missed nothing. People will make very bad data selection data choices to satisfy their preconceived idea of what the results should look like. Briffa’s silence on the selection criteria is deafening. Briffa knows the study is garbage, we know it is garbage, the folks at RC know it is garbage, and everyone else knows it is garbage… all thanks to Steve McIntyre’s efforts in getting the original data released. The only difference is in the willingness of different people to admit the study is in fact garbage.

  105. SteveF–
    I”m sure the guys at RC think Briffa was not garbage. Mind you, their opinion about the study may be mistaken, but i’m sure they do not think it’s garbage.

  106. Lucia,

    Well, I didn’t say they would admit it was garbage. If they really do not think it is garbage, then they haven’t been critically evaluating the study. No scientist or engineer (without an axe to grid) could think Briffa 2000 is anything but garbage. It has a whole series of flaws that make the paper’s conclusions unsupportable.

  107. Lucia
    There is a whole literature on pre-screening too. I think the preferred “technical” term is data-snooping or data-peeking. Steve MacIntyre (need I say!) had some good discussions on all this as well as references on Climate Audit. I might see if I can dig out the link.

  108. lucia,

    Briffa 2000 is in quality a lot like the studies we have previously discussed that claim to “prove”, using climate models, that 10 year periods without warming should be “quite common”. Of course, nobody at RC is going to say that these studies are garbage either, but I think we (you and I at least, plus a lot of other people) agree that they are.

  109. Lucia,

    Just had a quick look at the econometric references link at CA and didn’t find data snooping, but there is this rather good article on data mining in

    economhttp://www.climateaudit.org/pdf/others/datamining.greene.pdfics

  111. Unfortunately most of the links to the papers don’t sem to work. But you can get teh falvour from teh following abstract

    “Economics is primarily a non-experimental science. Typically, we cannot generate new data sets on which to test hypotheses independently of the data that may have led to a particular theory. The common practice of using the same data set to formulate and test hypotheses introduces data-snooping biases that, if not accounted for, invalidate the assumptions underlying classical statistical inference. A striking example of a datadriven discovery is the presence of calendar effects in stock returns. There appears to be very substantial evidence of systematic abnormal stock returns related to the day of the week, the week of the month, the month of the year, the turn of the month, holidays, and so forth. However, this evidence has largely been considered without accounting for the intensive search preceding it. In this paper we use 100 years of daily data and a new bootstrap procedure that allows us to explicitly measure the distortions in statistical inference induced by data-snooping. We find that although nominal P-values of individual calendar rules are extremely significant, once evaluated in the context of the full universe from which such rules were drawn, calendar effects no longer remain significant.”

    From SULLIVAN, R., A. TIMMERMANN and H. WHITE, 1998. The dangers of data-driven inference: the case of calendar effects in stock returns

  112. The people at RC don’t think the study is garbage. They have started to believe the MWP period did not exist. So any data to the contrary just means the data is flawed.

  113. MikeN,

    “The people at RC don’t think the study is garbage.”

    I would hope they are smarter than that, but you may be right.

  114. Lucia,
    R is very easy to learn as is Matlab. While Matlab may not be free, there are ways of getting it for free 😉 I have to say that R is much better than Matlab for the calculations that I’m doing. Remember I was always using Matlab a long time ago? I started using R because there was so much that I wanted to do with Matlab but couldn’t because it is a massive memory hog. It takes a very very long time to start up compared to R. Since R is a lightweight program, I can actually load large amounts of gridded data in R whereas Matlab would freeze for a minute and then spit out an out of memory error.

  115. Lucia,

    I understand that your model is specifically created so that no temperature extraction is possible:

    “Generate roughly 148 years worth of monthly “tree – ring” data using rand() in EXCEL.”

    By generating random data, you are making the assumption from the start that there is no correlation between tree rings and temperatures – you have created a wood of 154 trees that don’t agree with each other, with no particular correlation for one temperature history over another.

    Of course then if trees are selected from this wood to try to reconstruct a temperature history, that history will be a result of cherry picking, and will represent nothing.

    But your model is not a valid comparison to reality, and hence the conclusions you draw from the model have limited applicability. There are large observed correlations between tree rings – that was how it was recognised that such rings reflected growth conditions in the first place. There hence is a signal to extract, even if it takes some work to separate it from the noise.

    There are three ways to quantify that such a signal is there in the first place: firstly, what are the internal correlations between the tree-ring data, before even looking at the temperature data? This will tell you whether tree rings are likely to be measuring anything. Such a test would immediately show that the tree rings of the wood you had randomly created cannot be used to determine any influence, be it temperature, precipitation or improved forestry management.

    Second, once you use a correlation with temperature to exclude trees, what proportion are you throwing away and what is the correlation of those you include. In your case you exclude a huge proportion (how many of your 154 trees are left for your first plot?), which again tells you something is wrong – you don’t have a responding wood.

    Once a chronology is created from a particular series, there is a third test – how does it compare to the chronology from other series? Again, in your case there will be no correlation, but that is not what is observed in dendroclimatology.

    All three are valid tests of real data and don’t represent cherry picking. In comparison your random tree-ring model proves very little unless you make the initial assumption that tree rings are unable to measure temperature.

  116. TomP,

    Lorax, Nick Stokes, and others have actually claimed, on various blogs, that it is only logical that you would only use trees that “validated” against the temperature record implying that Briffa’s work ending up with 1 extremely important core and a few others was OK. They offered little on how to determine if the trees could actually BE thermometers or whether Briffa should even have done this. They also rejected, or ignored, the idea that this test of thermometerness should be performed on a different set of trees and the results used to screen the actual target trees BEFORE performing VALIDATION.

    Now, based on your long post I would suggest you run around the Blogosphere and straighten these gentlemen out who appear to be suggesting doing in reality what Lucia did as an exercise in what NOT to do.

    Of course, before you do that, you appear to have jumped from a correlation between tree rings and growth to trees being thermometers yourself with arm waving. How would you suggest this screening be done??

    You also claim that three tests are actually used successfully in DendroClimatology. Well, would you please give us links to Briffa’s documenting this in his paper?? Having this same information on the papers depending on Mann and Briffa papers would also be really nice. I am told that his paper appears to not contain this information!! We would love to know his actual selection and testing procedures to exclude the non-treemometers and their results and the post verification and results!!

    You do realise that even after you edited your comment you are still castigating Lucia for showing what SHOULDN’T be done, and why, and pretty much discussing the same points you list???

  117. I like these simple simulation examples. It’s far to easy to imagine that one can visualise the effects which are being discussed here, and ignore the possibility of doing some modelling with simulated data. This post gave me enough of an understanding to read Jeff Id’s (linked in the ~10th comment) which is very powerful, and introduces two more points:
    * Over responders (in the selection period) are more likely to be selected. (because the selection period has a slope)
    * Since the calibration period includes correlated noise, the reconstruction will reduce the scale of the variations in the reconstruction period.

    Coming back to intuition, if we select proxies which are well correlated, wouldn’t we expect the reconstruction to over-predict the signal in the calibration period?

    I was not sure about Jeff’s iso-temp lines, but am coming to the conclusion that the calibration period bias and error bars CAN be estimated for these series.

  118. The short answer is you can’t just cull data this way. It’s an error.

    If you know what your data are supposed to measure (and generally they do), you throw away data points that don’t fit with what you “know” to be the right answer. That’s cherry picking at it’s core regardless of why you choose to cull out part of the original data sample.

    So what do we do with the Bristlecone pines?

  119. Tom P,

    Just because the tree ring growth is correlated between trees does not make them temperature proxies. It could be some other common factor or it could simply arise from autocorrelation at individual tree level.
    The point of Lucia’s example is precisely that to even get off the starting block your reconstruction needs to outperform Lucia-type simulations. Steve Macintyre showed that MBH98 could not do taht, becaue tehy had not adequaltey allowed for autocorrelation.
    And you say that the proxies are correlated with each other. This is misleading. Of course they are correlated in the instrumental period, because that is how they have been selected. But before the instrumental period you will see a whole variety of behaviour. Just look at some of the spaghetti graphs. The lack of coherence in series before the instrumental period strongly suggests that the confidence intervals for temperature estimates from multi-proxy reconstructions if properly constructed would be huge.

    Finally, the best test of whether your temperature proxies are the real deal instead of over-fitted nonsense is how the equation forecasts on new, out of sample, data. The killer blow for me is that the proxies do badly – hence the divergence problem.

  120. kuhnkat (Comment#21804)

    “You also claim that three tests are actually used successfully in DendroClimatology. Well, would you please give us links to Briffa’s documenting this in his paper?? Having this same information on the papers depending on Mann and Briffa papers would also be really nice. I am told that his paper appears to not contain this information!! ”

    Don’t believe all you’ve been told – indeed isn’t that the true sceptic position?

    For example, please read Briffa in Phil. Trans. R. Soc. B 12 July 2008 vol. 363 no. 1501 2269-2282 at:

    http://rstb.royalsocietypublishing.org/content/363/1501/2269.full

    in which internal series correlation, correlation to temperature and correlation between series are all discussed and quantified.

  121. From the link of Tomp “”This is achieved empirically by aligning and averaging measured ring widths from all available samples by relative tree age (assuming in this case that the first sample ring represented the first year of the tree’s lifespan, and making no allowance for assumed difference from the true germination year) and using an age-related smoothing of these data (Melvin et al. 2007) to provide a practical reference curve. Each sample ring-width series can then be expressed as a series of deviations from this curve by dividing measured by expected values for the appropriate ring age of tree, in effect producing a series of relative indices that are then averaged, with respect to correct calendar age, to produce a dimensionless time series or chronology. This so-called regional curve standardized (RCS) chronology preserves short- and long-term variability of tree growth but mitigates against spurious changes in the mean chronology arising from the temporal coincidence of young-tree (relative wide rings) and old-tree (narrow rings) samples (Briffa et al. 1992 and earlier references therein). “”

    Jeffid is this what you found? I thought the valient 12 did not meet this criteria.

  122. TomP

    By generating random data, you are making the assumption from the start that there is no correlation between tree rings and temperatures

    Yes. The point is to show you get these sorts or results when there is true correlation.

    Showing the range of results one gets when the data really contain no signal is standard practice in statistics. If the result you attribute to “signal” is indistinguishable from highly probable results you get when there is no “signal”, your result is not statistically significant. If you don’t understand this point, then you will not understand why screening is a big problem.

  123. There are known ways not to datamine (cherrypicking) for biological studies. One states the criteria, say larch on a northern slope. One states no worm blight, no rotten cores, must be greater than 100 years old, how many cores per tree, etc. Next one takes these areas that meet the criteria, and cut & paste to a square grid until covered. A computer randomizer, gaussian, draws circles at an X:Y coordinate and numbers of trees to be taken in each circle. The area is determined by an X:Y coordinate on the square paper. The computer gives randomized identity numbers, since some will be thrown out in the lab. Extra sites and extra cores are kept to replace the thrown our samples, and ensure enough out of sample are kept to be run with the in-sample to confirm or reject the findings. ETC, ETC… This is what we biologists are supposed to do in cases like this. What Briffa, etc have done is to take the data screening part where you determine north slope, no worms, etc, and have not done the in-situ confirmation. THis is shown by using series rather than using random selection of sites and trees in the series. They continually and continuosly data snoop. Methods to prevent this go back to the 1930’s. It is not that these papers are wrong, it is that they are bad science. If their contention is correct, they have done enough studies to select the criteria, they have enough sites, they do the methodology and report the results. If it cannot be done a priori in this manner with CI’s, p-values, etc, it is merely speculation. Just as currently, what is offered is simply data snooping, and correlation = causation non-science. The fancy stats are like logic, a sure way to be wrong with the utmost confidence that one is right. If these series proxies are correct, a random selction of all from that study, if there are enough (Briffa 12 fail this criteria) one could take the present archives and do this. However, the amount of metadata would be tremendous as would be the number of samples required due to the number of known possible effects to tree ring diameter and density. In a believable study, these confounding factors would need to be shown inappropriate by falsification along the lines of verifying trees are thermometers. Some could be replaced by “fancy” stat procedures. But it has to be remmebered that each of these requires more discrimination of data, or relationships, that simply may not exist. Which brings me to a final point, If a procedure, say RCS, states the need of hundreds and one has 12, no one should be defending a misapplicationof a procedure. That and it are simply wrong.

  124. TomP

    Don’t believe all you’ve been told – indeed isn’t that the true sceptic position?

    For example, please read Briffa in Phil. Trans. R. Soc. B 12 July 2008 vol. 363 no. 1501 2269-2282 at:

    Actually, I dislike it when people do not summarize what they think the paper says in their own words.

    Basically: Could you tell us how you think they showed their method works, why you think it’s right and why we should believe their results are right?

    Otherwise, we have arguments like: “Here’s a paper.” “That paper’s crap!” “It’s peer reviewed.” “It’s still crap!!” No one learns anything.

  125. lucia (Comment#21812)

    “Showing the range of results one gets when the data really contain no signal is standard practice in statistics. If the result you attribute to “signal” is indistinguishable from highly probable results you get when there is no “signal”, your result is not statistically significant.”

    True. But looking at your constructed data:

    1. You had no correlation in the initial data.

    2. You had to throw away most of the data during cherrypicking.

    3. You saw no correlation between the various results produced after cherrypicking.

    If actual tree-ring chronologies exhibited any of these three characteristics, there would indeed be a profound problem with the entire approach. But it doesn’t – the actual data is immediately distinguishable from your model data.

    To put it another way, if you sent your complete model data to a dendrochronologist for consideration they would immediately say that they couldn’t use it as there was no signal of any kind apparent in the data.

    Even if you sent them the subset that correlated with temperature, the dendrochronologist would be highly suspicious as there would be no correlation in the pre-instrument record.

    And if you sent them several series that had no correlation between them, your results would again be recognised as unusable.

    It is precisely this ability of time-series correlation to exclude noise from signal that is the basis of many branches of science.

  126. lucia (Comment#21817):

    “Actually, I dislike it when people do not summarize what they think the paper says in their own words.”

    So why snip off my summary underneath? I wrote: “in which internal series correlation, correlation to temperature and correlation between series are all discussed and quantified.”

    To be exact:

    1. Internal series correlation: see the Rbar figures in Table 1
    2. Correlation to temperature: see figure 6
    3. Correlation between series: see figures 8 and 9

  127. TomP

    True. But looking at your constructed data:

    1. You had no correlation in the initial data.

    Of course not. That’s the standard test, and you just agreed. So why make this a but?

    2. You had to throw away most of the data during cherrypicking.

    Yes. And if you read the discussion of the “weak test” above, you’ll see we discuss that the quantity of randomly selected data that passes or fails could be a test of whether or not there is any signal in the set of data.

    If we were discussing a particular paper, we could talk about whether or not the data in tha paper would have even passed that test. But we aren’t talking about any particular paper, just the general notion– as The Lorax and Hu discussed it. That is: Generally. As a general notion, this is an issue, and you have to check.

    The purpose of the exercise in this post is to highlight the general issue so that people agree that– as a general issue– it exists and causes this problem.

    It is precisely this ability of time-series correlation to exclude noise from signal that is the basis of many branches of science.

    Erhmm… no. If we wanted to do more complicated things,we could show that you will oversate the level of certainty in your results by some amount if you screen this way. This will always happen if you screen this way. The amount

    So, the researcher has to a) admit this is an issue and b) do analyses to adjust his confidence intervals for having done this. Depending on the properties of the data you might find that your results have no significance– or you might find there is significance. But if the researcher hasn’t adjusted or checked for the effect of screening, then for all we know, the result he got is nothing more than the result forced onto the data by screening.

    I’m not saying this applies to any particular paper. I don’t know. There are zillions. It sounds like it applies to some and not others.

  128. Tom P (Comment#21821)

    I asked this:

    Basically: Could you tell us how you think they showed their method works, why you think it’s right and why we should believe their results are right?

    I snipped the rest because it would be an example of not showing how you think they showed the method works and etc.

    I did not ask for a list of page numbers. I want you to write your own words explaining what you think they claim and explaining why you think what they did is valid based on your own understanding of statistics.

    So, in other words, you tell us what you think is Briffa proved, showed or argued on those page numbers? (Like… pretend this is your thesis defense, or a final in an undergraduate class. You know…. don’t just say “It’s discussed in Smith. Period.”

  129. TomP–
    For what it’s worth, if I used screened data as in this post, I could concoct figures just like those for the screened data!

    So, none of that shows screening is not a problem.

  130. Tom P

    You say that proxies move together which suggests they are measuring something real. There are formal tests for this sort of coherence, which McIntyre has deployed for some of the reconstructions. In

    http://www.climateaudit.org/?p=3379

    he discusses the inconsistency statistic suggested by Philip Brown for just these calibration problems. For MBH99 he finds that the consistency statistic drops in the instrumental period compared to the pre-instrumental period, where it is often insignificant. In other words the series are coherent (correlated) in the instrumental period, but diverge from each other much more frequently in the pre-instrumental period. If the series were really thermometers then this would not happen. Much the most likely explanation seems to me to be that the coherence in the instrumental period is explained by data mining, and the big reduction in coherence outside the instrumental period is explained by the other non-temperature factors which impinge on the proxies. since there is no way of sorting out all the factors the idea that extrapolation of the proxies tells us something about temperatures is lust plain wrong.

  131. TomP:

    You had no correlation in the initial data

    That demonstrates that the problem with experimental bias is worse not better, since even the “best case” of experimental measurement error with no bias still produces a net bias during the calibration period.

    Obviously you can start with a distribution that has a correlation in it (e.g., see PaulM’s code above), and you still get the same effect, it just requires less tuning.

    2. You had to throw away most of the data during cherrypicking.
    3. You saw no correlation between the various results produced after cherrypicking.

    Which wasn’t the point of Lucia’s Monte Carlo simulation, so it’s not surprising that it doesn’t satisfy 2 or 3.

    If you use use the random process given in PaulM’s code above, that starts with you end up with about a 40% retention value, and a net correlation in final result. If we are allowed to flip signs, we end up with about 80% data retention.

    None of this should have been necessary to point out to anybody with an experimental science background.

    Link to modified version of PaulM’s code allowing us to reflect the sign.

  132. Tom, how confident are you of the correlation after ‘cherry-picking’?
    Did you run correlation numbers for the Yamal trees? The historical period trees do not appear to be that well correlated as you suggest. Then again that series was not run thru a cherry-picking algorithm.

    Mann’s results took something like 400 out of 1200 proxies as correlated. Is that a good number in your opinion?

  133. TomP,

    Just looked at the paper and do not see how to pull up the addendums.

    Where are the full procedures, calculations, verification stastistics…??

    I do not see what you claim is in this paper. Could be I am not as smart as I think. Please quote paragraphs and what you think they mean.

  134. TomP,

    This recent paper of Briffa in no way can be used to support his earlier paper. Please tell us the procedures, calculations, and stastistics for the Briffa (2000) paper that is referenced not only by this recent paper but also by other Hockey Stick papers.

    You claim these tests are common in Dendroclimatology yet this data, which we do NOT know how it was developed, is being used in Key papers!!

    I won’t bother you with more than a mention of how common Mannian abuses are also.

  135. MikeN

    Mann’s results took something like 400 out of 1200 proxies as correlated. Is that a good number in your opinion?

    Interesting. That means he threw out roughly the same fraction I threw out in my example. I didn’t happen to know that.

  136. Tom wrote the following:
    ———-
    True. But looking at your constructed data:

    1. You had no correlation in the initial data.
    2. You had to throw away most of the data during cherrypicking.
    3. You saw no correlation between the various results produced after cherrypicking.

    If actual tree-ring chronologies exhibited any of these three characteristics, there would indeed be a profound problem with the entire approach. But it doesn’t – the actual data is immediately distinguishable from your model data.

    ————
    I did this post a long time ago and it answers all of these criticisms. It’s the same link as above. Turnkey R code is provided so you can do it yourself.
    http://noconsensus.wordpress.com/2009/06/20/hockey-stick-cps-revisited-part-1/

    Criticisms by number:
    1 – It uses actual Mann data.
    2 – Mann 08 throws away most of the data – there are other similar papers which do the same thing through selective weighting.
    3 – They are primarily tree ring proxies from the same sources as dendros use so we can be certain to find the correlation you require

    So since actual tree ring chronologies DO exhibit the same problems Lucia has outlined, you of course agree with the point that Mann 08 is garbage.

  137. Jeff Id (Comment#21833)

    http://noconsensus.wordpress.c…..ed-part-1/

    Looking at your post, I’m not surprised you can construct a range of reconstructions by using a correlation to an predefined signal by using a low proportion of the dataset. But with only 6, 18 or 10% of proxies retained these reconstructions are hardly representative of the dataset.

    Of all the reconstructions you show, though, it looks like the calibration signal that retained the highest proportion of proxies, at 42%, was in fact the signal that best matched the modern temperature record. This seems to be a good reason to choose this result over the others you reconstructed.

    In any event, Mann’s paper is using a range of proxies, not just tree-ring data, so it’s not obvious what is being excluded through correlation with the temperature record.

  138. @TomP, This seems to be the view I had of these investigations until this morning. Jeff’s part 1 strikes me a little bit as a reconstruction based on an arbitrary set of non-orthogonal waveforms. Interesting, but not conclusive.
    Read this one http://noconsensus.wordpress.com/2009/06/23/histori-hockey-stick-pt-2/ and it will hopefully help you to realise that the selection of noisy proxies based on correlation will tend to prefer those proxies where the noise adds to the correlation, and suppress those where the noise makes the correlation worse. This has the effect of making the selected proxies seem more sensitive to the signal than they really are. In the example of part 2, the proxies DO correlate well in the calibration period, and the final result is still wrong.

  139. Nice job Tom, you picked the right answer, that was Mann’s excuse as well. However we shouldn’t forget Mann pasted (regressed) a hockey stick blade on the end of over 95% of the proxies and used temperature series for 71 of the Luterbacher version.

    In many cases, the pasted on blade wasn’t strong enough. In fact by picking the right criteria I retained 39% of the proxies when comparing to a downslope.

    http://noconsensus.wordpress.com/2008/10/11/will-the-real-hockey-stick-please-stand-up/

    It’s kind of fun having someone to try and criticize this nobody else has tried.

    Perhaps now you agree that Mann08 is rubbish.

  140. Pingback: How to trick yourself into unintentional cherry picking to make hockey sticks « Watts Up With That?

  141. Jeff Id (Comment#21840):

    Looks like you thresholded at a much lower correlation than Mann to get 39% of the data to match an artificial recent cooling signal in:

    http://noconsensus.wordpress.c…..-stand-up/

    I’m afraid this post consists of lots of plots signifying nothing.

    In the supplement of Mann 08 the following point is made:

    “Although 484 (~40%) pass the temperature screening process
    over the full (1850 –1995) calibration interval, one would expect
    that no more than ~150 (~13%) of the proxy series would pass
    the screening procedure described above by chance alone. This
    observation indicates that selection bias, although potentially
    problematic when employing screened predictors… does not appear a significant problem in our case.”

    If you’re serious about questioning Mann’s paper, this would be the line of attack, not the random plots you’ve created which are themselves below the inclusion percentage and instrumental correlation of the very reconstruction you’re criticising.

  142. Sorry to propose a perhaps banal thought.

    I found really nice and illuminating this discussion on the problems and pitfalls of proxy screening,
    but was thinking of an even simpler – and admittedly much more banal – recipe for HS manifacturing.

    Step 1: you have or build a group of “non-proxies” – say tree rings – that you wrongly assume to be temperature proxies for the last millennium.

    Step 2: you accumulate and average a lot of them and, as they, for your purpose, contain only noise you get an horizontal line: this is the handle…

    Step 3. Then, for the last period in which temperatures are known and have increased,
    because you want to “screen the good treemometers” you select only the rising pseudo-proxies, or, that is the same, you deselect enough non-rising ones:
    this reproduces “correctly” the blade.

    Could this explain a large percentage of the topic?

  143. JeffId did you paste (regress) “”a hockey stick blade on the end of over 95% of the proxies”” and use “”temperature series for 71 of the Luterbacher version?”” to yours and see if you get ~40% pass the temperature screening process? Or does the screening occur before the pasting?

  144. So, Hu is correct. If you screen out “bad” proxies based on a match to the current temperature record, you bias your answer. Given the appearance of the thermometer record during the 20th century, you bias toward hockey sticks!

    The same applies to models. Who would present model that doesn’t hindcast. i.e Show the known (sic) temperature record? None. The reason being anyone with a modicum of thought would say we can’t make forecast if you can’t get the ‘known’ facts right. It’s common in fianance, and its called a selection bias.

    That means you quite often end up with models that are nothing more than a curve fit.

    Nick

  145. Tom P

    Looks like you thresholded at a much lower correlation than Mann to get 39% of the data to match an artificial recent cooling signal in:

    It’s not fair to you because I’ve spent too much time with this paper. However, here’s a quote from the paper.

    significance thresholds are r  0.11 and r  0.34, respectively.

    The 0.11 is the primary cutoff threshold – by far I think only like 2 or 3 of 484 used the higher r. However in agreement with your point slightly more series are retained in the other ‘hockey stick’ direction.

    Now in my graph at the link provided:

    ‘cps pattern match r=0.1 series percent used = 39.43.’

    So I think you’ll find your statement about the level of my threshold to be somewhat inaccurate. However, in your favor I’m not claiming it’s exactly equal. The stats are more complex than that but you can see that I made no effort to exaggerate the situation.

    Now perhaps you’ll agree Mann08 is rubbish.

  146. John,

    I used the same pasted series with temperature data and sorted the ones which correlated best to a downslope and still was able to find 39% which could pass the very low criteria of r=0.1.

    The edit feature is cool but it’s causing me a bit of trouble in that it doesn’t display.

    My point was that I used the actual post-pasting data from Mann08. The HS blade wasn’t strong enough to overcome every series and I still got a very strong downslope when compared to an r of 0.1 which is basically the same as Mann used for his paper.

    It was a bit surprising that a negative slope would correlate to that level with so many series. Against TomP’s criticism that is basically the same value used in Mann08.

    This isn’t a minor point WRT the paper.

  147. I should add that even if the 39 percent was only 3 percent, my criticisms of the paper wouldn’t change.

  148. Tom P, quoting Michael Mann:

    This observation indicates that selection bias, although potentially
    problematic when employing screened predictors… does not appear a significant problem in our case.

    Mann says his methods are OK. Wow, what a relief to know that the issue is settled… How persuasive.

    And so we can also dismiss Jeff Id’s detailed, documented offerings as “lots of plots signifying nothing” without actually having to explain why they ‘signify nothing’ because …well just because.

    Mann found a screening method that (a) preserved the tree ring studies done by his friends and some suspect Finnish lake sediments while (b) eliminating hundreds of other data sets that would not have fit the hockey stick plot. Pure, objective science at work. Golly, I’ll bet he was just as surprised as the rest of us when the hockey stick once again emerged from the data selection…

  149. “Golly, I’ll bet he was just as surprised as the rest of us when the hockey stick once again emerged from the data selection…”

    I heard he was like a kid opening a birthday present.

    “Honey…time to open presents!”

    (eyes closed and wincing)

    PLEASEBAHOCKEYSTICK
    PLEASEBAHOCKEYSTICK
    PLEASEBAHOCKEYSTICK
    PLEASEBAHOCKEYSTICK…

    😉

    Andrew

  150. TomP said:
    “All three are valid tests of real data and don’t represent cherry picking. In comparison your random tree-ring model proves very little unless you make the initial assumption that tree rings are unable to measure temperature”.

    And that is the point, Tree Rings cannot be used to measure temperature. There are too many variables to absolutley say that a tree ring is a measure of only temperature!

  151. Thanks for this really clear explanation, Lucia. Any chance you can put up your spreadsheet so we can use it as a base to try our own experiments?

    Cheers!

  152. “Deep Climate” over at “Real Climate” is upset about this thread, complains vaguely about lucia’s treatment of the subject without saying where she’s wrong. Thought I’d take a look. I don’t see the problem, Deep. Well, there is one teensy problem. No cherry-picking thread is complete without the Esper quote, which was – get ready for this – published in a peer-reviewed article.

  153. The Esper quote, and my parsing of it, can be found at:
    http://cruelmistress.wordpress.com/2009/10/13/cherry-ping-pon/#comment-698
    I’ve given up on Ben Hale’s blog. But there was a terrific exchange there with “luminous beauty”. The Enlightened One, so far, refuses to debate me at Climate Audit. Maybe he will show up here? He has a lot to say to people like lucia.
    .
    Ah yes, the quote:
    “The ability to pick and choose which samples to use is an advantage unique to dendroclimatology.”

    Thanks to Steve McIntyre for unearthing this beauty.

  155. Why reject the uncorrelated at all?

    It seems a bit presumptuous, if not hubristic, to say any portion of the sample can be rejected given our limited understanding of the processes involved and the tens of trillions of dollars of policy being guided by these studies. A signal from the full sample is suspect enough.

    It seems like basic conservative elements of the scientific method are just being thrown out here for the sake of ideology.

  156. Bugs,

    “McIntyre is quite happy to exclude bristlecone pines.”

    Due to his extensive auditing and study he has good reason.

    Now, if you would like to see what the NAS panel actually wrote read Chapter 4 Page 52 here:

    http://www.nap.edu/catalog.php?record_id=11676

    Surface Temperature Reconstructions for the Last 2,000 Years

    (hint strip-bark = bristlecones)

    The rest of it is worth a read also.

  157. Don’t forget JAcoby’s original explanation of cherry-picking – when he refused to produce the results for the 26 series of 36 collected that he did not use in their NH composite – see http://www.climateaudit.org/?p=29

    The inquiry is not asking for the data used in the paper (which is available), they are asking for the data that we did not use. We have received several requests of this sort and I guess it is time to provide a full explanation of our operating system to try to bring the question to closure…

    If we get a good climatic story from a chronology, we write a paper using it. That is our funded mission. It does not make sense to expend efforts on marginal or poor data and it is a waste of funding agency and taxpayer dollars. The rejected data are set aside and not archived. ..

    As we progress through the years from one computer medium to another, the unused data may be neglected. Some [researchers] feel that if you gather enough data and n approaches infinity, all noise will cancel out and a true signal will come through. That is not true. I maintain that one should not add data without signal. It only increases error bars and obscures signal…

    As an ex- marine I refer to the concept of a few good men.

    Jacoby was the mentor to Rosanne D’Arrigo who explained to ta surprised NAS panel in 2006 that you have to “pick cherries if you want to make cherry pie”. No kidding.

  158. bugs:

    McIntyre is quite happy to exclude bristlecone pines.

    And I don’t include measurements from sensors not sensitive to signals in the frequency band I’m interested in studying, when analyzing data.

    I guess to your sloppy notion of “cherry picking”, that would cherry picking too.

    Only to bugs is it ok to use the data being fit to to decide which proxies to include, but one shouldn’t perform any screenings not including correlating with the signal of interest to determine which tree species make good temperature proxies and which don’t.

  159. Jeff Id (Comment#21846)

    From the supplementary text of the Mann paper it is apparent that different r screening thresholds were used for the annual and decadal data. There was also screening in the two calibration periods. I can’t see in your R code where this is implemented.

    One good test to see if you are implementing the same method as in the paper is to reproduce the results. Have you done this?

  160. TomP is frankly engaged in magical thinking here that somehow the particular signal-proxy screening protocol that Mann is using is going to protect him from the effects of cherry picking, when in fact there is no a priori expectation that it would do so.

    If TomP thinks that some how Mann’s ad hoc methodologies have protected him from the effects of cherry picking, given the strong evidence to the contrary, it’s up to Mann or to his white knight TomP to demonstrate this now.

  161. Carrick (Comment#21865)

    If there was just cherry picking going on in Mann’s 2008 paper, as per Lucia’s model of random data, you would expect only 13% inclusion. In fact 40% of the data passes the screening procedure, which indicates that there is a substantial signal.

    Could you please produce “the strong evidence to the contrary” here?

  162. “McIntyre is quite happy to exclude bristlecone pines.”

    Due to his extensive auditing and study he has good reason.

    So McIntyre can have good reason, but the no one else can. Got it.

  163. So this is where Tom P has run off to. Tom, you made a claim on CA about the corelation of the Yamal chronology to the local temperature grid. namely that it was above .4 to a very high confidence. I asked you to produce the chart and code to support this position. You ran away ( last I checked.. maybe you posted it.. ) anyways people here pretty much know what I think about claims without supporting data and code. As much as I have disagreed with nick stokes and arthur smith those two guys had the character to make their claims and back them up and stick around for the whole show. Tom P, you have an unfinished report to turn in for Dr. bender and Dr RomanM. I suggest that you finish that if you want any credibility in the circles where lukewarmers travel

  164. Jeff Id (Comment#21846) October 17th, 2009 at 5:44 pm

    tom is in denial. There are two forms of denialism. there is the denialism of ignorance and willfull denialism. Tom P practices the latter. When you demonstrate that he is wrong he will

    1. change the topic
    2. flee the thread.

    mark my words.

  165. Lucia and Jeff,

    I imagine you’re right, but to convince more people you need a simpler test. One idea would be to see what shape you get if the calibration period is 1940-1970, when thermometer temperatures didn’t move much. No doubt the blade of the hockey stick disappears. This will show that you only get an uptick at the end from proxy data when you impose this by calibrating those data against an uptick. Apologies that my resources at present do not allow me to do this for you.

    Another point that might be worth making in public discussion is that steady reversion to the mean as you follow the proxy curve back in time is actually evidence of a poor reconstruction. We know temperatures move noticeably with little or no greenhouse forcing – you can see it in thermometer records from 1910 to 1940 and you can see it in some proxy reconstructions other than Mann’s. So a practically straight line with only minor wiggles going back 1000 years until it reaches almost exactly the midpoint temperature of the correlation period is a strong hint that the method is useless and that the proxies it selected are just noise.

  166. tom is in denial. There are two forms of denialism. there is the denialism of ignorance and willfull denialism. Tom P practices the latter. When you demonstrate that he is wrong he will

    1. change the topic
    2. flee the thread.

    mark my words.

    My irony meter just broke.

  167. TomP,

    The screening thresholds are quite simple in practice. The results of the different methods are provided in the excel table on his site. Mann to his credit was quite open in presenting his results this time.

    You now you ask if I have replicated Mann’s work. The answer is no I have not, as I specifically point out in the links I provided. This code is a simplified demonstration of the problems of CPS such that the open minded can understand what exactly Mann did to make a hockey stick from 1209 trees and other assorted thermometer and non-thermometer items.

    If you want the full code replication Steve McIntyre has it on CA full R code provided for your perusal. Anyway, I didn’t expect to convince you of anything including the color of the sky.

    So you said to Lucia that her method is wrong because the data is random. I showed real data with the same results. You said, not enough data retained so it’s not real. Surprisingly I showed about the same percent retained when matched to a downslope. Now you ask have you replicated the full paper and the statistical results. The full replication has already been done by SteveM – code provided of course. But I wonder, if I waste my time to reproduce the flawed verification that would absolutely PASS this inverted reconstruction as valid will you then admit Mann08 is rubbish?

    Dave Brewer,

    It’s a good suggestion actually. Shorter time periods pass acceptance more easily but why not. BTW, at the links I’ve left above I’ve got some R code which can be easily tweaked to do your experiment. It’s very thoroughly commented, turnkey and designed to be messed with like that.

    SteveM’s Jacoby quote is priceless. It’s sad really because the famous Jacobi was very good at math, this Jacoby isn’t doing any good for the name.

  168. David–

    but to convince more people you need a simpler test

    Convince people of what? I have repeatedly said this post is not about any particular paper. At Climate Audit, there was an argument about whether or not this effect could happen at all. The argument wasn’t about any specific paper but whether or not the issue is even worth thinking about in principle.

    If we are to go by TomP’s quote, Mann admits there is an effect. So, why people like Lorax, TomP or others start out by rejecting it out of hand is mystifying.

    As for using a different calibration periods: Of course the blade will disappear if you use 1910-1940 as the calibration period. I don’t even need to run the test to know that. The method a) forces the average to take on the shape of the calibration period (no matter what that shape is and b) revert to a fairly flat smooth curve outside the calibration period.

    So, finding a blade in papers that use 1910-1940 as the calibration is more convincing than finding it in paper that use the later part of the 20th century as a calibration. Screening still has problems– but screening using 1910-1940 won’t force a sharp blade on the end It tends to flatten anything outside the calibration period. (It might cause step jumps so the average pre-1910 is different from post 1940. But that’s not “blade-like”.)

    All this examples shows is: When someone screens proxies with the temperature record, we need to be very, very careful about evaluating that. (And of course, since the screening might happen prior to entering cores into an archive, we need to be worry about that too.)

  169. SteveM

    If we get a good climatic story from a chronology, we write a paper using it. That is our funded mission. It does not make sense to expend efforts on marginal or poor data and it is a waste of funding agency and taxpayer dollars. The rejected data are set aside and not archived. ..

    The italicized is a very, very bad practice. It introduces “pre-screeing” into the archive. After that, no scientists can know whether his signals are due to screening or random noise. And the only statistical tests possible must be applied on a decent run of temperatures outside the thermometer record that existed prior to archiving. (That would mean that even the slightest divergence problem with data after the core was archived needs to be taken very, very seriously! That’s the only “out of sample” test! )

  170. Lucia,

    At the end of reading this long thread, what do I find at the bottom? An ad for Cherry trees!

    “A Cherry Tree
    Find Tip, Coupons and Deals at
    CouponMoutain.com”

    Is Coupon Mountain associated with RealClimate? Just askin’…

  172. JeffId,

    “But I wonder, if I waste my time to reproduce the flawed verification that would absolutely PASS this inverted reconstruction as valid will you then admit Mann08 is rubbish?”

    Don’t waste your time. It seems TomP’s belief in Mann08 and similar proxy studies is not rational. It won’t matter how he responds to your question, and it won’t matter what the analysis shows, he will never accept that many proxy studies are profoundly flawed (as you say, pure rubbish.. or as I would say, pure crap) due to data selection biases.

  174. Actually, lucia, selective calibration on the early part of the instrumental period (say, 1900-1940) is still bad. That is a period with a modest warm trend and it is just as problematic for getting inflated calibration statistics (especially if you’re not archiving the misfits!). You will produce “hockey sticks” with low MWP and high CWP. So avoiding the late instrumental period and period of divergence (post-1950) does NOT solve the problem. There are lies out there in the blogosphere to the contrary. luminous beauty being one of the perpretrators. But Gavin’s new guru – Deep Climate – is suggesting same.

  175. The hockey stick inflation factor is the calculated correlation during the calibration period divided by the true (and unknown) correlation during the reconstruction period. If they are equal, no HS bias added. Anything that inflates the numerator will increase HS bias. Some like to pretend that the denominator is known (or can be guessed at from insight?), but it is not.

  176. Distortion of the archive is a real issue. Gaspe was one of the key series in MBH98 – it had a huge HS and propped up certain robustness tests. The series ended in 1982. I’d been to Gaspe in the late 1960s and 1970s on business. During our research, I learned that there was an updated version of Gaspe which had never been archived or published. I asked for a copy of the data and D’Arrigo/Jacoby refused to provide it, saying that the original HS version provided a better signal.

    I asked where they had sampled and they said that it was before GPS and they didn’t know for sure. I’ve dealt with geologists going to far more remote places than Gaspe and they locate things on maps. And you can be sure that there was an appropriate scale forest service map.

    We discussed the location over the next couple of years and I think that we did eventually figure it out from a travel itinerary posted by a tourist on the internet.

  177. bugs:

    So McIntyre can have good reason, but the no one else can. Got it.

    Nobody is arguing with selecting for specific species, site locations and so forth that are thought to lead to go proxy samples, and selecting against others.

    All dendrologists do that.

    What we’re talking about is something different, which is the simpleton error of using the correlation between data and proxy to select which samples to use in the final average. That’s an error whether you are Steve McIntyre, bugs, Carrick, Michael Mann, or (gasp!) even Lucia.

    Lucia discussed above using the correlation between proxy and data to identify certain characteristics of the specimen that may have made it a better proxy than others that didn’t show a good correlation.

    Then once you’re done with that, acquire new samples with these characteristics, and produce a temperature reconstruction with just these new samples.

    Again there is nothing wrong with that.

    Got it?

  178. bender–
    To clarify…. You mean using the 1910-1940 is bad if there proxies are “signal+noise” right? I would agree. You might exaggerate.

    On the other hand, I’m pretty sure if I used my “pure noise” the 1910-1940, I wouldn’t great the sharp upblade we see in this post.

    It would be easy enough to “tweak” the spread sheet to put, say, 20% signal in the pseudo-proxies and see what happens. (To those asking for the spreadsheet, I’m going to put it up. I’ll discuss this 1910-1940 issue tomorrow.)

  179. Bugs,

    “So McIntyre can have good reason, but the no one else can. Got it.”

    My statement was followed by a link to a report that a board set up by the NAS wrote. I gave you the Chapter and Page. Here is the quote you apparently could not be bothered to read:

    While “strip-bark” samples should be avoided for temperature reconstructions, attention should also be paid to the confounding effects of anthropogenic nitrogen deposition (Vitousek et al. 1997), since the nutrient conditions of the soil determine wood growth response to increased atmospheric CO2 (Kostiainen et al. 2004).

    Your statement totally misrepresents what I wrote.

    Try again.

  181. “So few papers.”

    Boris, time to update your System Date and Time. It ain’t yesteryear anymore. This is the Information Age. The Appeal to Papers is not only a logical fallacy, it’s an outdated logical fallacy.

    Andrew

  182. Boris–
    You do realize that if trees don’t contain sufficient signal to make a reconstruction that fact limits the number of correct papers that can be published? Generally speaking, peer reviewers tend to favor papers saying a) “I tried X, found Y and what I found Y is statistically significant” rather than “I tried X and
    b) discovered there was no signal statistically significant information in X”.

    Since there is also a fairly well known tendency for journals to not really like comments or direct rebuttals, the result is you can get papers like (a) and then criticisms of (a). (Quite honestly, that observation is not even controversial. In most fields, bad papers are just ignored by nearly everyone except the authors and their co-authors and grad students.)

    Historically, people would share opinions at conferences and still do. Now there are blog- criticisms are tending to appear on blogs. So, claiming a paper is “good” because the criticisms are only found on blogs is … well… a bit much.

    If ti’s good, you should be able to explain why it’s good on the papers own merits. If you can’t do that, that means you don’t know. (Which proves little other than you don’t have sufficient expertise. But that doesn’t mean others who do have expertise can’t comment on the papers. They will.

    This may be a shame, but the real fix is for journals to become more liberal about encouraging negative comments, permitting them to be as long as the original papers (and sometimes longer). What might actually be better is for journal to not publish negative comments, but instead support an interal archive of the pds and provide a lists of all negative comments received in the journal and provide a link to the comments so that subscribers can read what those comments and judge. This would be easy to do these days.

  183. BTW, we had a frost here in Kentucky this morning. I was tempted to pour some coffee in my Global Lukewarming Mug, but I’m saving it for a beer the day I hear Big Al say he doesn’t believe in Global Warming anymore. 😉

    Andrew

  185. Boris–
    Me? Or people?

    Yes. Researchers know that it’s more work to put criticism in journals. So, in many fields, they budget their time and efforts to tasks that have greater pay-off. Like, for example, spending their time working on their funded projects, writing proposals to get funding and writing their own positive papers. So, over all, there is less “trying” to publish criticisms.

    Sort of like there is less banging heads against walls rather than building nice new houses. The one is futile, and hurts. The other has a payoff.

  186. Lucia,

    “What might actually be better is for journal to not publish negative comments, but instead support an interal archive of the pds and provide a lists of all negative comments received in the journal and provide a link to the comments so that subscribers can read what those comments and judge. This would be easy to do these days.”

    I agree that it wold be easy to do, but I don’t really think this is going to happen, for a couple of reasons. First, there is no “peer review” of comments, so no “quality control”, and no editorial protection of published papers; I don’t think editors would ever accept this. Second, it costs a bunch of money to publish a paper, but it would cost nothing but your time to make a negative comment (unless publication fees were involved in making comments); I suspect that journal editors would see this as an unfair advantage for commenters vis-a-vis authors.

  187. TomP:

    If there was just cherry picking going on in Mann’s 2008 paper, as per Lucia’s model of random data, you would expect only 13% inclusion. In fact 40% of the data passes the screening procedure, which indicates that there is a substantial signal.

    Lucia’s model used uncorrelated random noise. This says, even if you have uncorrelated measurement noise, you will still get a 13% inclusion of non-signal components into your measurement. A 3 to 1 signal-to-noise ratio is “very poor” not and not a “substantial signal.”

    However, as I pointed out earlier, using PaulMs correlated noise model, you also get about 40% “data” passing the screening model. That doesn’t look that great for Mann’s methodology to me.

  188. SteveF– s
    Sure. It would give an unfair advantage to commenters— sort of. This assume people would take comment as seriously as the actual peer reviewed papers. Also, I’m not sure there would be a flood of comments. Even if the comments just appeared as un-reviewed comments, many active researchers would be might take care, knowing that being undiplomatic can hurt you later when the person you criticized acts as reviewer.

  189. Lucia,

    But what would keep someone, say “Luciene” or “Seven McIntosh”, who do not publish in the dendro field, from commenting harshly about Dr. Brifano’s inexplicable tree selection protocol or his complete lack of archived raw data? I think editors will always want some mechanism to screen comments and protect the published papers.

    Perhaps some current or former journal editor could comment.

  190. SteveF–
    Nothing would stop that. Right now, they editors over-screen (even if they don’t think they do.) With some brainstorming, they could move toward less screening. I’m not sure they will want to do this, so I doubt they would spend much time doing it. But, in the meantime, blogs will exist, and some will get attention.

    Right now, editors can do nothing to prevent SteveM from getting attention. He gets it. People can say “not peer reviewed” all they want. But peer review isn’t really that ancient of a practice, and science made perfectly good progress before it existed. It will continue to do so even if people like “Luciene” or “Seven McIntosh” were able to post comments at journals themselves.

    Scientists would either a) read the comments or b) not read them. You would likely find that these comments would be read by some scientists, they would chat with each other and, if maintained in a permanent archive, some would be cited and some not. Same as peer reviewed articles. Same as grey literature has always been treated. This should not be a big deal.

    Do I think it will happen? Actually, I think in 20 years you’ll see something like this. Or… given technology, maybe 5 years. But the 5 years will be in fields that are less politically charged. 🙂

  191. “Me? Or people?”

    People.

    “Sort of like there is less banging heads against walls rather than building nice new houses. The one is futile, and hurts. The other has a payoff.”

    I agree, but some people are already doing the work–or most of it–and already butting heads. I wonder what the payoff is in all that?

  192. Boris–
    If publishing articles is NOT a consideration for promotion at your job, there can be much more psychic pay off in having discussion with people at blogs than trying to submit comments in the comment and reply cycle and having going through the various hoops.

    Admittedly, this route might not be one taken by many who are beavering away for tenure, where the only thing that counts is peer review articles (as it might actually be counter productive). But people blog about knitting, cooking, photography, gardening, their kids etc. Why wouldn’t some blog their evaluation of journal articles of scientific arguments for exactly the same “pay off”?

  193. Boris has perhaps never heard the oft-told coffee-room joke about “The Journal of Negative Results”. There’s a reason why it’s tragicomedic.

  194. bugs (Comment#21874)

    Bugs in case you missed the oft repeated memo I believe in AGW. I also think that the denialism practiced by those who should know better ( defenders of Mann) is more pernicious more dangerous and more akin to holocaust denial than the denialism of AGW. Frankly I can stomach the uneducated who don’t believe in radiative physics. They can always be educated, but when people who ARE educated, who have studied science, when those people deny such simple things as the data and code must be available, when they deny that geographical coordinates should be published accurately, when they refuse to archive data, when they defend even trivial errors to the death, and when they refuse to use the name of a critic, then I think, this is something rarely seen in history. Educated knowledgeable people ignoring almost willfully what they know to be true .. en mass.

  195. Boris (Comment#21893) October 18th, 2009 at 9:44 am

    That’s a interesting truth test. Do you count the papers for and against? do you weigh them? count the words? count the number of readers? count the editors? the peer reviewers? count the degrees of the peer reviewers? here is a simple test. Dr. Mann published a paper wherein he misplaced the coordinates of a proxy site. This is something that you yourself can verify. He was alerted to this mistake. he repeated the error. You will find no paper published that corrects this error. No errata that corrects it. Therefore, it must not be an error. But Mann himself did correct the error, without notice or attribution. It’s not in any published paper. what to make of that?

    http://www.climateaudit.org/?p=4303

  196. Jeff Id (Comment#21876)

    “This code is a simplified demonstration of the problems of CPS such that the open minded can understand what exactly Mann did…”

    Your code demonstrates there could be problem, and one that is recognised in the field, but it does not do what Mann did.

    “… I didn’t expect to convince you of anything…”

    I would be convinced if code was presented that duplicated Mann’s results and showed an alternative reconstruction that met all the screening criteria of his paper. Until then it’s rather difficult to conclude that the paper is “rubbish”.

    steven mosher (Comment#21871)

    It’s a little strange seeing you drag your obsession with me onto another site. I gave you some time ago the reference to the correlation between the tree-ring and instrument data above 0.4 published in Briffa’s 2008 Proc. Roy. Soc. paper. If you want a code to produce that result I suggest your time might be more healthily spent learning R rather than indulging in this obsessive behaviour.

    Carrick (Comment#21900)

    Paul’s M code is another way of illustrating Lucia’s point. I do not dispute that hockey sticks can be artificially constructed. What I object to is that such a construction could be done without anyone knowing it – in fact some care is taken to guard against such a selection bias by Mann in his 2008 paper.

    PaulM and Lucia use artificial data while Jeff Id uses real data but does not follow the screening procedure of Mann. If Mann is wrong it really shouldn’t be so hard to prove it along the lines I suggested to Jeff Id above.

  197. Tom P doesn’t understand the quantitative argument – and has been avoiding it at CA: if your trees have a weak signal plus lots of noise (say, correlation = 0.2) and you increase the apparent signal strength – either by ignoring divergent samples, or screening out entire chronologies – you may boost your apparent signal to a correlation of 0.4 or even higher. A doubling of the apparent calibration over the real calibration leads to a 2x hockey stick inflation factor. Because you’re inflating teh correlations in the instrumental period without doing so in the historical reconstruction period.
    .
    lucia’s toy model works on white noise and no signal and so requires severe screening. The example I’ve given above is based on red noise and weak singal and requires very little screenging. ESPECIALLY in a population of divergent samples. Her model illustrates perfectly why Lorax is wrong and Hu is right. My model explains why Briffa and Esper are wrong to practice what they apparently do: undocumented screening.

  198. Tom P:
    Do you know why Jeff Id can not do exactly what Mann did? Because Mann’s code is secret. The only reason you can play this game at all is because Steve McIntyre empowered you by releasing his code. Get the irony here?

  199. Tom P says:
    “some care is taken to guard against such a selection bias by Mann in his 2008 paper”
    Yeah, right. Like including the bristlecone pines in the SI Fig 8? Go to CA and read where Nick Stokes is working the same dodge you are. You blew your cred at CA and are now pushing your misunderstanding into other parts of the blogosphere. Congratulations. You do a massive disservice to the alarmist movement, who desperately need help from skilled people very much unlike yourself.

  200. TomP

    I would be convinced if code was presented that duplicated Mann’s results and showed an alternative reconstruction that met all the screening criteria of his paper.

    All this statement shows is that you do not understand the problem. That’s fine. Your ignorance of know statistical problems means you will not be convinced by correct arguments and will instead be convinced by incorrect ones. In some sense that’s fine with me. I don’t try to convince ignorant people who believe in astrology they are wrong. But, I think they are wrong and I’ll say so.

    Now, to repeat: some of Manns screening criteria can, in some cases, create hockey sticks out of nothing. So, of course no one is going to apply those to anything and create a something with no hockey stick. This doesn’t prove is methods are wrong or that the right answer may not be a hockey stick. But what I know is this: The evidence you point to does not present a good argument that his method did not create the hockey stick out of nothing. (Now, it may be that something else in that paper does. But it’s clear you do not know enough about statistics to recognize which things would prove that his paper showed what he claims.)

    Second, if the proxies do not contain sufficient information, no one can show an “alternative reconstruction” that has any statistical significance. In this case, you would could have two types of papers: People who don’t understand statistics who think they’ve found something and people who do understand statistics who know there is nothing to report! The fact that you insist on people claiming they have created a reconstruction when it’s possible that the data just aren’t enough to show anything simply demonstrates your ignorance.

  201. Tom P says”
    “If Mann is wrong it really shouldn’t be so hard to prove it along the lines I suggested to Jeff Id above.”
    The flaw in Mann’s work is that he relied on the Graybill chronology which NAS said clearly “do not use”. Mann thumbs his nose at NAS, and you think Jeff Id needs to follow your directions to disprove the guy’s work? You are in as much a state of denial as Nick Stokes. Remove Graybill and invert Tiljander and all of a sudden things aren’t so “unprecedented” anymore.

  202. TomP,

    ” If Mann is wrong it really shouldn’t be so hard to prove it along the lines I suggested to Jeff Id above.”

    Then you should have no problem doing this and throwing it in our faces.

    When should I check back to see your reproduction of Mann ’08????

  203. Lucia,

    as a cat lover, could I impose on you to host a thread of TomP’s reproduction of Mann ’08, if he does it??

  204. bender (Comment#21911)

    “…you’re inflating teh correlations in the instrumental period without doing so in the historical reconstruction period.”

    But if this were indeed the case we would expect the agreement between the reconstructions to break down prior to the instrument record. That is why I specified the third test that there should be agreement here to ensure there really is a climate signal to be seen.

    Such agreement is observed in the published literature. For instance from the Mann 08 paper:

    “…all of the individual reconstructions that pass validation fall within the uncertainties of the composite reconstruction, defined as the average of all individual reconstructions that pass validation. In other words, the various reconstructions are consistent within uncertainties.”

  206. “That’s a interesting truth test. Do you count the papers for and against? do you weigh them? count the words? count the number of readers? count the editors? the peer reviewers? count the degrees of the peer reviewers?”

    C’mon, even Wegman said that blogs were inappropriate for discussing these issues. And why they are inappropriate should be obvious.

  207. lucia (Comment#21914)

    “The evidence you point to does not present a good argument that his method did not create the hockey stick out of nothing.”

    Please then explain what is incorrect in this statement from Mann’s paper:

    “Although 484 (~40%) pass the temperature screening process
    over the full (1850 –1995) calibration interval, one would expect
    that no more than ~150 (~13%) of the proxy series would pass
    the screening procedure described above by chance alone. This
    observation indicates that selection bias, although potentially
    problematic when employing screened predictors… does not appear a significant problem in our case.”

    In particular, as your statistics appears to be more advanced than mine, you might want to check if the one-sided significance thresholds would indeed screen for the number given of random proxies.

    You say:
    “The fact that you insist on people claiming they have created a reconstruction when it’s possible that the data just aren’t enough to show anything simply demonstrates your ignorance.”

    No, I’m saying something very different: if you think there is a problem with say Mann’s paper, it does need to be demonstrated.
    It’s not enough to say there is a possible problem. There is a burden of proof quite reasonably placed on anyone who makes an accusation against a published paper.

    I actually don’t think putting together alternative constructions is the best way to show it, although Jeff Id had already embarked along that route. But there would certainly be something to report as a rejoinder if Jeff showed an alternative, admittedly false, construction that was compatible with the paper’s screening procedure.

  208. “even Wegman said”

    Appeal to Wegman. Logical Fallacy.

    Because someone named Wegman has opinion A, doesn’t make him correct.

    Andrew

  209. “There is a burden of proof quite reasonably placed on anyone who makes an accusation against a published paper”

    Appealing to the god named Published Papers again. 😉

    Andrew

  210. kuhnkat (Comment#21916)

    “Then you should have no problem doing this and throwing it in our faces.”

    So it’s not up to the critics of Mann to prove him wrong, it’s up to me to prove him right!

    The concept of “Onus probandi”, that the burden of proof rests with those that make the accusation is important in science as well as in law.

  211. Oh common Boris,
    Blogs aren’t an inappropriate forum for discussing topics. They have disadvantages that specialists in the field are not going to spend time reading all the blogs because it’s not time efficient. But that doesn’t make it inappropriate for you, me, Tom, Andres, Bender etc. to post our reactions. That notion is just silly.

  212. TomP
    Simple. This conclusion is not fully supported by the evidence:

    This
    observation indicates that selection bias, although potentially
    problematic when employing screened predictors… does not appear a significant problem in our case.”

    Mann would have to show more to support the conclusion.

    In particular, as your statistics appears to be more advanced than mine, you might want to check if the one-sided significance thresholds would indeed screen for the number given of random proxies.

    At what level of significance? And the way Mann does it? Or the way it should be done?

  213. “There is a burden of proof quite reasonably placed on anyone who makes an accusation against a published paper”

    Burden of proof to whom?

    As a generality, this statement is just hooey. The fact that a paper was published– and/or was approved of by about 3 people does not catapult it to some stellar level of truth where everyone in the world is required to accept it no matter how obviously nonesensical. People are allowed to say it’s flawed, provide their reasons. Others who under stand the flaw are permitted to recognize the proof and accept it.

    Of course, this doesn’t mean you have to accept it. You also are not required to believe Gerlich and Tennahauser (sp?) between the time it was published and anyone published a formal rebuttal. I mean… sheesh!

  214. I’ve just realised where I had seen this before: its curve fitting 101. Any random data set of m points can be fitted over a portion of the set (t points < m points) by a polynominal function or trend of power n where n can be 1 to t -1. Now in this case each proxy could be fitted with a different polynominal over the calibration period such that the average result approaches zero trend. But as you point out, you are screening the polynomials to how well they match the shape of the temperature curve which then biases the results. Any subset of the random data set will always have a net trend (as the other remaining set, its conjugate, will have the opposite) and as the filter is not specific about which set to choose you are performing a combination comparison rather than a permuation exercise, selecting the best components of the subset to meet the criteria. And its a great example you have chosen and demonstrated
    To get around this what should happen is that another independent property of the tree that can be matched to temperature, if it exists, should be used as a control and if the situation exists that this property has fidelity with ring width, those tree rings can be used as a proxy. Otherwise the data, even if there is a signal, can't be used as the method doesn't distinguish true signal from noise. In atomic physics for example the very low output signal ( orders of magnitude amplitude below average white noise amplitude) is chopped (modulated) because the input signal is chopped. The output can then be identified as being a result of stimulus and isn't noise, then it is mode-locked and amplified. If it wasn't modulated then it would just look like noise.
    Basically then two bits of information are needed to justify using a tree as a thermometer.

  215. Tom P

    The answer to your question is that Mann’s statement should not be taken at face value. He has
    1)included instrumental readings among his proxies
    2)used proxies whihc have been cut off at 1960 becasue after taht they diverge
    3)used the infamous upside-down lake varve site from Finland the 20th century portion of which shows thick varves because of human intervention not temperature
    4)uses the bristlecone pines which are known not to be good temperature proxies
    and
    5) does not appear to allow for autocorrelation when calculating the claimed confidence levels, thus not really testing against the possibility that all he is picking up is red noise.

    And his pre-instrumental coherence tests look ad hoc and nothing like as rigorous as those suggested by Philip Brown for calibration problems.

  216. “That notion is just silly.”

    The notion that blogs could ever resolve an issue is the silly notion. Blogs are about stroking egos and picking fights.

  217. Tom P,

    You may force me to take that last step. It isn’t very difficult though. During the early months after the paper was released one of the questions that was examined was this:

    you might want to check if the one-sided significance thresholds would indeed screen for the number given of random proxies.

    I believe it was Dr. McCulloch who investigated this and found that due to insufficient corrections for autocorrelation, the number of proxies which would pass in random data was something like 30%. It’s just from memory so I could be wrong about the person (maybe it was Roman) and exact amount but it is what I remember now. The number was substantially larger than reported but before Lucia would be able to confirm anything there are a number of different proxies with different levels of autocorrelation to deal with – In Mann 08 it was solved with a substantial underestimation.

    I already knew of this so that’s why I said:

    reproduce the flawed verification that would absolutely PASS this inverted reconstruction

    I already know it would pass verification by random data and the underestimated autocorrelation but I have a second advantage. Since the verification portion is based on a central portion of the signal and the signal sorting is based on an upslope between the beginning of calibration and the end, the central ‘verification’ portion can hardly be considered independent. The center would of course be forced to have an upslope. That’s worth a post by itself. Maybe later this week.

  220. Boris–
    Resolve? Who said resolve? What you said was discuss. Here’s the quote

    C’mon, even Wegman said that blogs were inappropriate for discussing these issues. And why they are inappropriate should be obvious.

    Individual journal articles rarely “resolve” an issue. When they do, other issues generally come up, so something is alway unresolved.

  221. “Blogs are about stroking egos and picking fights.”

    Sometimes they are. But often they are used as a way for people to actually communicate with each other. Real people can pass on information to other real people and back again. It is a place where information is accessible. This is a good thing. And it is probably the reason YOU are here, Boris… because of the information exchange.

    Or are you here just to propagandize?

    Andrew

  222. TomP,

    “So it’s not up to the critics of Mann to prove him wrong, it’s up to me to prove him right!”

    Wrong. You did not understand my comment correctly, or, you understood it and are trying to weasel out.

    “If Mann is wrong it really shouldn’t be so hard to prove it along the lines I suggested to Jeff Id above.”

    Prove it isn’t hard to do along the lines you suggested to Jeff Id above.

    The fact that you start arguing about the issue tends to make me think you can not do it.

  223. Because of the lesson it teaches, that lesson’s importance, and clarity — this is one of the better “non fiction” blog posts I’ve read.

    Congratulations Lucia and your like minded contributors to this thread.

  224. Tom P (Comment#21921)

    You use the following quote from Mann to somehow indicate that Mann is not subject to the problem that started this thread:

    “Although 484 (~40%) pass the temperature screening process
    over the full (1850 –1995) calibration interval, one would expect
    that no more than ~150 (~13%) of the proxy series would pass
    the screening procedure described above by chance alone. This
    observation indicates that selection bias, although potentially
    problematic when employing screened predictors… does not appear a significant problem in our case.”

    However, you seem unaware that the Mann quote does not help your cause. What the above Mann quote actually shows is this (and only this):

    1. A collection of stationary AR(1) red noise series with autocorrelation coefficients similar to proxy data would pass screening approximately 13% of the time.

    2. The proxies passed it 40% of the time.

    And that’s it. It doesn’t show anything else (regardless of what you, Gavin, or Mann may want it to imply). It doesn’t show that selection bias is not a problem – the only way to show that would be to not screen anything, run it through the Mannomatic, and see what results. It doesn’t even show that the proxies contain signal and not noise, because the noise series used to arrive at the 13% may not be the appropriate series to use. If the real noise model should turn out to be something like an ARFIMA model, the noise may pass as frequently or more frequently than the proxies.

    Not only that, but it still doesn’t show that the proxy “signal” is really a temperature signal. Since temperature and CO2 are fairly collinear during the calibration period, the proxy “signal” could easily be CO2. Or precipitation. Or some weird combination that is dependent on location, species, and even the specific tree.

  225. Lucia,

    For some reason the previous post lists itself as being by “Tom P” instead of me. Not sure how that happened. Not sure if you can clean it up. Haha. 🙂

  226. TomP:

    What I object to is that such a construction could be done without anyone knowing it – in fact some care is taken to guard against such a selection bias by Mann in his 2008 paper.

    That’s an opinion, not an argument, and from somebody, pardon me for saying, who has absolutely no experimental research background. So you are not exactly an authority as to what does or does not constitute a reasonable effort to prevent selection bias.

    Although 484 (~40%) pass the temperature screening process over the full (1850 –1995) calibration interval, one would expect that no more than ~150 (~13%) of the proxy series would pass the screening procedure described above by chance alone

    That’s because he’s not using a realistic noise model.

    Do it right, and I believe you’ll find virtually no difference in terms of rejection percentage for signal versus signal+noise, if you give the noise the correct stochastic properties of the tree ring samples. (Note that’s a different thing than saying you can’t perform an analysis that could differentiate signal from noise, just not using the approach Mann is using.)

  227. Ryan O:

    Not only that, but it still doesn’t show that the proxy “signal” is really a temperature signal.

    This is well put.

    Simply getting a correlation over a narrow time region doesn’t prove that the proxy continues to correspond to temperature outside of the calibration interval.

  228. Tom P, you have no business continuing to claim significance for Mann’s proxies. Please read the relevant posts at CA – for example, http://www.climateaudit.org/?p=4908 4216 3821 3838 3858.

    If something is a temperature “proxy”, then it should have a significant correlation in both “early” and “late” periods as defined in M08. Only 342 do (and one of these has a “significant” correlation with a different sign in each period, thus only 341.

    Further breaking down the 342 of 1209: 71 of 71 Luterbacher proxies pass – these use instrumental data and thus are irrelevant. It is irresponsible of you (and Mann) to inflate the statistics by including the high correlation of instrumental temperature with instrumental temperature as evidence of proxy significance. C’mon.

    Of the 927 dendro proxies with code 9000 (i.e. excluding the already RegEM’ed MXD data), only 143 (15.4%) pass the first-cut test, barely above Mann’s 13%. In fact, it is probably no better than random if Mann’s benchmark were done properly. There are two things left out in Mann’s benchmark, both of which would reduce the apparent “significance”: first autocorrelation, as others have observed; second, Mann’s weird pick-two method, where he uses the higher of two correlations in the observation, but doesn’t allow for this in the benchmarking. As I observed to you, I didn’t try to re-do this benchmarking at the time, but it is evident that the benchmarks will be even higher, squeezing the already unimpressive yield of dendro proxies. I might try to do it some time, but in the meantime, I can reasonably observe that Mann’s benchmarking is incorrect.

    On Mann’s criteria, all the code 9000 proxies should be eliminated because they are no better than random.

    The MXD proxies were tortured in a variety of ways: Briffa’s MXD network was made into gridcells by Rutherford, Mann et al 2005 using RegEM; in addition, they truncated the divergent real data at 1960 and infilled data after 1960 using other proxies, perhaps including instrumental data. 93 out of 150 of these heavily massaged series are “significant” – much higher than non-RegEMed non-infilled dendro data. It’s not a small job sorting out exactly what they did, as it requires sorting out Rutherford, Mann 2005 RegEM which is different than subsequent RegEM.

    As you say, Mann and coauthors undoubtedly thought “very carefully” about getting to their claim of ~40%. Unfortunately, that means that readers also have to examine their claims “very carefully” – something that you once again, have failed to do.

  229. TomP,

    “The concept of “Onus probandi”, that the burden of proof rests with those that make the accusation is important in science as well as in law.”

    Except Mann et al never release enough information to duplicate their results, even when that information is formally requested, so anybody that doubts the conclusions (or even if they just want to confirm them) is in a difficult position. Mann (as usual) says “I did an analysis, the results are like this…trust me”. Surely you can see that withholding data and analysis methods is inappropriate.

  232. As you say, Mann and coauthors undoubtedly thought “very carefully” about getting to their claim of ~40%. Unfortunately, that means that readers also have to examine their claims “very carefully” – something that you once again, have failed to do.

    They have every right to publish what they want, you are effectively vandalising the long established and highly successful scientific method. If you think you are right, publish. Be as careful as you want,.

  233. Boris…
    Huh?
    Jeff wrote

    My own estimate is Mann08 = 100% bias.

    How does his posting his estimate transform into a claim that blogs can resolve scientific issues?

    Jeff is participating in a discussion at a blog. He posts his estimate. Obviously this doesn’t “resolve” the issue for everyone on the planet and he didn’t claim it did. Publishing a peer reviewed paper wouldn’t “resolve” the issue. People discuss results of peer reviwed papers. That’s what’s supposed to happen!

  234. Bugs,
    .
    “you are effectively vandalising the long established and highly successful scientific method.”

    Please… that is preposterous.

    Scientific publication and the scientific method are completely separate. I can (and do) use the scientific method (hypothesis, experiment, results, analysis, conclusion, repeat the cycle) all the time, but rarely publish my work, since most of what I work on is not of broad interest. The problem with certain published research in climate science is that it is of very broad interest and also demonstrably flawed. It is the people who publish nonsense, and who then hide their data and analysis methods, who corrupt and vandalize the “highly successful scientific method”.

  235. “you are effectively vandalising the long established and highly successful scientific method”

    van·dal·ize

    TRANSITIVE VERB:
    van·dal·ized , van·dal·iz·ing , van·dal·iz·es

    To destroy or deface (public or private property) willfully or maliciously.

    He’s malevolently destroying the scientific method? In what way? How? That claim doesn’t even make sense.

    And he’s doing it effectively? All by himself?

    Pfffft 😉

    Andrew

  236. Nice Boris – how about “ok you have a point” next time. Or was that the main detail you were going to stake your rep on?
    ———
    I forgot about the pick two Steve mentioned. If one cell didn’t correlate well they chose another and tried again. Of course that affects the threshold of what’s considered random.

    Say you are applying for a job on the psychic friends network. You know you can predict events but must pass a test of your psychic ability to get the job. The interview involves coin tosses. Most of us have a 50/50 chance of flipping a coin and getting heads and you do it a 1000 times so the expected number of correctly guessed heads or tails for the non-gifted is 500.

    You know the data is noisy sometimes things don’t work out due to interference from local toasters. However, like trees are temp, you have real power your decision is to call heads or tails and then flip coin twice and record a positive result if either or both tosses come up heads. You repeat the process 1000 times. Then report on the application that your powers are confirmed and your 750 positive results proves your amazing power. You get the job and make $3/minute.

    Does anyone imagine that the authors involved with that portion of the paper didn’t realize that the statistical threshold for randomness might shift a little if they checked the correlation of the data twice. It took serious code to make it work but the important detail is that IF they didn’t know, why do it at all ??!!!! It’s self proving, they know exactly what they did. And then after they did it why not shift the asserted threshold for expected random correlation?

    Mann that’s ugly.
    —-
    A hundred percent Boris. IMHO.

  237. Pingback: Qohel - Australia's Leading Independent Conservative Blog

  238. “There is a burden of proof quite reasonably placed on anyone who makes an accusation against a published paper.”

    Well, Let’s say I grant you this. I don’t but let’s say I do.

    Stipulated: the burden of proof lies with the reader.

    Now let’s talk about discovery. If mann doesnt release EVERYTHING required to reproduce his work, all the data and all the code, then I would argue we can reject it OUT OF HAND. That is, we can rationally say he has not proved what he claimed to prove. And further it’s irrational to believe any scientific paper where the data and methods are not revealed.

  239. >So it’s not up to the critics of Mann to prove him wrong, it’s up to me to prove him right!

    Actually, yes Tom I agree with this statement. In general, the scientists are the ones making the case for global warming, and it is incumbent on them to prove their case.

    In the case of Mann and the hockey team, they are going against the previous consensus that was published in the first IPCC report. So the burden of proof is even more on them.

    As to this, I think if you search Steve’s site, you will find that he did replicate Mann08. It is this some code that Jeff is using. I asked Jeff about this a few months ago, when I was trying to do some different things, and wanted to start with a replication.
    It took me awhile to realize that the code has some differences. Namely, Mann08 compares proxies against local temperature, while Jeff is matching all of them to a single curve. So the question I was asking Jeff was pointless, when I asked for the function that produced the temperature curve.

  240. Stephen Mosher, I think Yamal is correlated to the local temperature record. It is about .39-.45, and Briffa has it at .56. I’ll eventually do a full evaluation of the correlations, but so far I got Schweingruber correlated a bit ahead of Yamal. I posted this at CA.

  241. Mike, it would be interesting to see how the insturmental correlation differences between Yamal and Schweingruber play out in the “CRU 12” period.

  242. Just one more point to add to Steve M’s list of criticisms of Mann significance levels. There is Lucia’s point that the data may have been screened even before Mann et al got hold of them. It is certainly far from obvious that teh original set is a random selection from possible proxies.

  243. Steve McIntyre (Comment#21944)

    “Tom P, you have no business continuing to claim significance for Mann’s proxies.”

    Why not? What are the qualifications for someone to comment on this? All I have done is read the Mann 08 paper and responded with what I think are the relevant parts of the text.

    “Please read the relevant posts at CA – for example, http://www.climateaudit.org/?p=4908”

    So you’re upset because I should have read your blog before commenting. Is this what is called the “relevant” literature? I have actually had a look here, but there is a much better way to make sure your criticisms are put on the record and read alongside Mann’s original paper.

    “If something is a temperature “proxy”, then it should have a significant correlation in both “early” and “late” periods as defined in M08. Only 342 do (and one of these has a “significant” correlation with a different sign in each period, thus only 341.”

    You said that in your postings. But how many of the 150 random proxies would pass this additional screening? The random proxies have to go through the same screening as the actual data before any comparison can be made here. Maybe this number is further down in the comments, but I can’t see it. This is an example of the problem inherent in making a scientific case on a blog.

    If a scientist wants their research to be considered by others in the field, they don’t generally refer to a blog posting, plus associated comments, as their work of record on the subject. I think its great to see blogging on scientific subjects. But if you think you have a case that overturns some published work, why leave it dispersed on a blog. Is there anything stopping you wrapping this all up and submitting it for publication?

  244. Bugs,
    To accuse Steve M of vandalising the scientific method is laughable. You would do better to turn your fire on people who do not publish their data and methods in a way that allows reproduction. What is your take on MBH 1998, for example, which said it used principal components analysis but in fact did not do so?

  245. MikeN (Comment#21958) October 18th, 2009 at 10:43 pm

    Thanks mikeN. At .39 it would fall below the cutoff that wilson suggests. Which thread did you post the code and graphs on? there are other criteria as well. Also, tom P suggested that dropping the last few years of Yamal would make no difference. Should be instructive

  247. Tom P, having the reconstructions agree with each other isn’t significant.
    Suppose I run 1200 random proxies(stock prices) through Mann’s algorithm, and a hockey stick pops out because the algorithm is flawed.

    Then I run a different 1200 proxies(popularity of names) through Mann’s algorithm and again a hockey stick pops out.

    Both of these will be very flat in the non-instrumental period, so they will be in agreement with each other.

  249. Didn’t Gavin acknowledge this pitfall and claim that it does not apply because the literature shows that the autocorrelation that you injected does not exist in the data? I personally think that his argument is ludicrous, but I remember him making it.

  250. “People discuss results of peer reviwed papers.”
    “people” don’t just discuss papers. They claim to have invalidated them. But they won’t man up and publish or write it up or contribute. I have no idea why you wouldn’t want to do that. Well, I do have ideas…

  251. Didn’t Lucia just explain to you Boris? Do you not understand English?

    And BTW, I read on all kinds of AGW blogs this argument that skeptics don’t publish. However, I’ve read many skeptic papers in publication and I’m aware of 2 reviews of Steve Mc’s papers receiving what are biased unfair reviews designed to prevent publication. It turns out you have to have a stick with a blade to talk about hockey. Even if your point is that the stick is bent.

  252. “They claim to have invalidated them. But they won’t man up and publish or write it up or contribute.”

    Boris,

    Your program is stuck in a loop. You need to change it so it asks:

    “Is the paper valid?”

    If it can be demonstrated that the paper is valid, the answer is “yes”.

    If it cannot be demonstrated that the paper is valid, then the answer is “no”.

    This can be done (as has been shown here and elsewhere) without obselete ceremony. You’ve seen it yourself.

    Andrew

  253. Boris,

    People discuss results of peer reviewed papers.”
    “people” don’t just discuss papers. They claim to have invalidated them.

    You think people at conferences don’t discuss papers? After papers are published either a) they are read and people discuss them or b) they are ignored and not discussed. If (b) they have no influence.

    You were concerned that blog discussions don’t “resolve” anything. Well, mere publication of a peer reviewed paper also doesn’t “resolve” anything.

    What the heck do you want to write a paper about? This post is about an issue discussed by The Lorax and Hu on a blog. It’s a well known problem. No matter how many papers contain this flaw, a paper sole on the flaw would be difficult to publish. It would also be difficult to publish a paper proving Archimedes principle. The reason is the reviewers would say “everyone already knows this.”

  254. Steve McIntyre (Comment#21966):

    “Tom P, you haven’t responded to 1) Mann’s inclusion of the Luterbacher proxies; 2) the non-significance of the 927 code-9000 dendro proxies.”

    Not my area, hence I’ve no immediate response. Do you have any answer to the number of random proxies which would pass the full screening? I presume this is in your area of expertise.

    You also haven’t responded to the the historical repeated divergence between younger and older trees in your combined Yamal-Khadyta chronology, despite your unsupported claim to the contrary at http://www.climateaudit.org/?p=7278.

    As this again very much falls in your area of expertise, why the reticence?

  255. MikeN (Comment#21965)

    “Both of these will be very flat in the non-instrumental period, so they will be in agreement with each other.”

    No, they would both be noisy and uncorrelated at all time scales before the instrument period. Please look at Lucia’s plots at the head of the post.

  256. TomP–

    Please look at Lucia’s plots at the head of the post.

    The plot at the head of the post doesn’t show individual proxies, so it will not tell you if they are correlated in the pre-thermometer period.

    That said: If the proxies were truly randomly generated, their noise will (mostly) be during the non-screened period. (There a little correlation can arise very near the approach of the screened period due to my having defined the baseline by the screened period and the need for all “good” proxies to conform to the temperature at the beginning of the baseline. But if you put a fairly long time buffer between the “screened” period and the “past”, you should be get a zero correlation for the past.

    This only means that screening this way isn’t what causes correlation in the past. The correlation in the past could be due to other factors. Certainly, we expect all trees on the same mountain to display correlation, but that’s not evidence supporting the “thermometer” theory. They could correlate because soil improved or degraded as a result of accumulation or degradation of humus in the soil, or maybe a disease hit a competing species of trees leaving room for the other trees to grow. (For example, the green ash borer is killing ashes right now. If a forest was 1/2 ashes and 1/2 oaks, and all the ashes die, the oaks will no longer have to compete with ashes afterwards. If the ash borer goes continent wide (as the dutch elm did before it) this could affect trees over a huge area, resulting in historic correlation.

    Could this happen in a climate reconstruction? Of course. Did it? I have no idea.

    When the paleo guys come up with reconstructions based on many different types of proxies a) using none upside down, b) not screening as in this post, c) using no proxies “upside down” and basically not resorting to anything suspicious in the published paper, people will stop criticising these proxies.

    I know you think somehow because a paper is peer revied, the “burden of proof” falls on the critics. But this is not so. The purpose of the paper is for the author to convince readers of his claim. If the readers are not convinced, then the author has failed at his burden of proof.

    (Mind you, he doesn’t have to convince everyone of everything. If his only goal is to get more funding, he may only need to convince a funding agency. Or other peer reviewers etc. But that doesn’t mean outsiders suddenly become convinced. They can doubt, criticize and say whatever the heck they want!)

  258. “No, they would both be noisy and uncorrelated at all time scales before the instrument period. Please look at Lucia’s plots at the head of the post.”

    Amazingly, Tom P still doesn’t seem to understand even the simplest point of the entire thread.
    Lucia’s blue plot is quite flat, and she only used 154 ‘proxies’. With 1200, it would be a lot flatter.
    For the benefit of those who know even less than I do about statistics, the noisiness of the flat part goes like 1/root(N).

  259. mikep (Comment#21962) October 19th, 2009 at 1:31 am

    Bugs,
    To accuse Steve M of vandalising the scientific method is laughable. You would do better to turn your fire on people who do not publish their data and methods in a way that allows reproduction.

    Reproducing results means get off your own **** and go out there and do the hard yards. It’s worked fine up till now.

  260. He has done the hard work and shown the results are pish.

    I would have thought holding onto data for 10 years and failing to allow others access is “vandalising the scientific method” and is laughable almost as laughable as those who try and defend such actions.

  261. PaulP:

    Amazingly, Tom P still doesn’t seem to understand even the simplest point of the entire thread.

    I accept this as evidence he isn’t interested in dialogue, because I haven’t seen him concede a single point yet. Rather he retrenches, shifts the goal posts, and tries again on another issue.

  262. Correct me if I am wrong, but isn’t it absolutely necessary to this technique that the data being used is random, and a particular type of random data, ie, red noise. Red noise is a random signal with a high autocorrelation (see step 1), which just means that successive values in the signal are always close in value compared to the permissable range of variance. Another name for red noise is “a random walk” or a “drunkards walk”. Because the signal is random, it will always over time revert to the mean. However, if (or when) it diverges from the mean, it will remain away from the mean for a substantial time because it cannot return to the mean in a short number of steps. As a result, if it diverges from the mean near the end of a run of sequences (either end) it will produce the charactersitc hockey stick shape. Average over a large number of random walks, the line will closely follow the mean over the entire length of the graph, because any areas of divergence from the mean in any particular run may be at any location in the run. If, on the other hand, you first rearrange the runs so a that any end region divergences are at one end only, and then use only the absolute value of divergences, and then take the mean, the mean will always have a hockey stick shape as a simple function of your data manipulation (this is what McIntyre and McKittrik did in their attempt to stitch up Mann et al.)

    Lucia does something different and only selects runs based on their correlation to an ideal end shape. However, here it is necessary that the runs be random if you want to guarantee a hockey stick shape. If instead, the run’s were sine waves with a randomly determined start point, then picking by correlation of the end curve would just pick out those curves in the same phase of the sine wave and your resulting graph would be a sine wave. If the curves were quadratic, this cherry picking technique would pick out quadratic curves rather than hockey sticks. It is only because all the curves from which Lucia picks were randomly generated that they all return to the mean. Had he instead picked randomly from a selection of curves, he would not have got his hockey stick.

    Here is the point: the growth of tree rings is known to be sensitive to certain climactic factors. There is a known causal link between rainfall, level of sunshine, temperature and even CO2 content and the growth of tree rings. Consequently the curve defined by the width of tree rings is not a random walk. What is more, it is known that tree rings are most sensitive in their growth to the factor in least supply. Trees in warm locations are not that sensitive to temperatures, in wet locations, not that sensitive to water, and so on. In very cold locations such as the edge of a treeline, they are very sensitive to temperature.

    So, if you pick tree ring data to match known temperature curves durring part of the end of growth, will you get a hockey stick? Possibly, but it is certainly not guaranteed. If the tree ring growth is temperature sensitive, and if the climate was much warmer in the recent past, the tree rings selected would show a large tick shape at the end. If the temperature followed a slow sine curve pattern as believed by many sceptics, the tree ring data would show the same pattern.

    If the tree rings were not sensitive to temperature, but only matched coincidentally, then they would follow the curve to which they where sensitive, be that rainfall patterns, summer sunshine, or what ever. There is no reason to believe a priori that any of these curves would be a hockey stick, so no reason to expect the hockey stick pattern.

    If you has a variety of each type, the genuinely sensitive to temperature,and the only apparently so, you would get the mean of a random selection of curves, not the mean of a randomly (red noise) determined curve, and again would have no reason to expect a hockey stick shape.

    So for Lucia to have a sensible criticism, all he needs to is show that tree ring growth is random, ie, it is not controlled by environmental factors.

  263. Tom Curtis:

    So for Lucia to have a sensible criticism, all he needs to is show that tree ring growth is random, ie, it is not controlled by environmental factors.

    Actually… not.

    The issue here is whether tree ring data acts as a proxy for global mean temperature, and not whether they are sensitive to environmental factors such as local temperature, rainfall, etc. Obviously it is true they are for those.

    Further, for many of these, they don’t actually have a monotonic relationship between tree-ring growth and these environmental factors… rather there is an optimal combination of temperatures and precipitation rates for which the tree will thrive, and for which you get optimal growth.

    Going from the global versus local variations, global temperature and local temperature or precipitation rates do not always track, often they anti-correlate for long periods of time.

    So, if you are going to posit a sensitivity to global mean temperature, these sensitivities to local environmental factors are going to appear as correlated measurement error in your reconstruction.

    The onus here really is on people like Mann to come up with something better than vigorous hand waving explanations for why they would ever have expected dendro growth patterns to be sensitive to global mean temperature to start with. (And I’m sorry, but invoking a buzzword like “telecommunications” isn’t a sufficient argument for this.)

  264. Lucia asks:
    “Could this happen in a climate reconstruction? Of course. Did it? I have no idea.”
    It did happen in the case of the Graybill bristlecone pine chronology. They knew the “sweetspot” was Sheep Mtn and they knew they could get an uptick in some percentage of trees. The cause for the uptick has never been explained, but it is not solely climatic. Moreover, no one can say to what extent Yamal larch are exhibiting the same inexplicable response. But you see a pattern here? Graybill versus Ababneh. Shiyatov vs. Schweingruber. Until you understand the source of the divergence in growth trends among trees (and among chronologies) you have no right screening out the divergent sub-population that does not match the climatic record. None whatsoever. Anyone who argues otherwise, I hope to see you in Washington, DC, when you take the witness stand.

  265. bugs says:
    “do the hard yards”

    bugs, do you have any idea why the Internatioal Tree Ring Data Bank exists? It’s because the paleo community recognized long ago that there was a need to share data globally if people were going to do global change biology. The community you purport defend does not share your monopolistic attitude. And neither do the policy-makers, who want quality science and transparent science. So it looks like you’re out on a limb – with your friend, Mann.

  266. Hmmm, BBC says new paper says tree growth better correlated to cosmic rays than temperature or precipitation…

    “The intensity of cosmic rays also correlates better with the changes in tree growth than any other climatological factor, such as varying levels of temperature or precipitation over the years.
    “The correlation between growth and cosmic rays was moderately high, but the correlation with the climatological variables was barely visible,” Ms Dengel told the BBC”

    http://news.bbc.co.uk/earth/hi/earth_news/newsid_8311000/8311373.stm

  267. Carrick says
    “The issue here is whether tree ring data acts as a proxy for global mean temperature, and not whether they are sensitive to environmental factors such as local temperature, rainfall, etc. ”
    .
    Just to put a finer point on it; it’s a little bit more subtle than this. Treeline tree growth is temperature limited, full stop. The non-zero correlation has turned up too many times for it to be entirely spurious. The issue is quantitative: how much is the correlation inflated by screening.
    .
    I know people like to simplify arguments to black-and-white. It’s useful for illustrative purposes. But doing so in this case tends to polarize a debate which is overly polarized as it is. Each side ends up arguing straw men that are not relevant.
    .
    What is the true correlation? What is the reported correlation after screening? As it is this difference that is the source of reconstruction bias, this is the way the question needs to be phrased. It’s quantitative.

  268. Steve might not like this, but in my opinion, he is Martin Luther, whose blog was a sheet of paper, hammer and nail, trying to force the reformation of an institution he dearly loves. His goal is not to destroy it, any more than Martin Luther wanted to destroy Christianity. He wanted to steer it back to its core values and fundamental power. This perversion of science, whether intentional or not, enables the sale of carbon indulgences, and makes the “pope” of AGW, and you all know who I am talking about, very wealthy.

  269. Tom P whines:

    “You also haven’t responded to the the historical repeated divergence between younger and older trees in your combined Yamal-Khadyta chronology, despite your unsupported claim to the contrary at http://www.climateaudit.org/?p=7278.

    As this again very much falls in your area of expertise, why the reticence?”

    First, a coherent argument is necessary. Several commenters at CA (a half dozen at least) have asked you to summarize your argument so that it is intelligible. Right now it is a diaspora of random complaints, graphs, and links. Second, do you really think – if you were to score a point here – that it would be game, set & match? C’mon Tom, be realistic. You’re making yourself look unecessarily silly. If you want to score your point, serve it up and let’s be done with it. Steve McIntyre is a squash player who knows when a point’s been scored on him. So stop goofing around and serve it up.

  270. You can take bender’s statement one mroe step.

    “Just to put a finer point on it; it’s a little bit more subtle than this. Treeline tree growth is temperature limited, full stop.”

    The further step is: Treelines move. This has been explicitly pointed out in the peer-reviewed publications involving the Yamal series.

  271. “I know people like to simplify arguments to black-and-white.”

    This is because every argument is either multiple true or false statements, or one true or false statement.

    If the argument is a collection of statements, it can be broken down into smaller pieces to see if the simple/fundamental parts are all true.

    A complex argument relies on all the simple pieces being accounted for.

    Andrew

  272. Tom Curtis (Comment#21981),
    “There is no reason to believe a priori that any of these curves would be a hockey stick, so no reason to expect the hockey stick pattern.”
    I think the non-temperature environmental signal may in many cases also be hockey stick shaped. A lot of things are influenced by human activity, which has a hockey stick shape. CO2 concentration, for example.

  273. lucia (Comment#21975)

    “The purpose of the paper is for the author to convince readers of his claim.”

    The burden of proof is on whoever is making a claim. When a paper is submitted for publication it has to pass the peer review process – the burden of proof is on the authors to make their case and respond to the reviewer’s criticisms. This may fix some more obvious problems with a paper, but of course some may remain, although the authors have refined what they believe at that time to be their best case in what they publish.

    Once the paper is published, it is up to the wider scientific community to criticise the paper. Science is essentially adversarial – it doesn’t matter how many people a paper convinces, it can only take one critic with a clear refutation to bring a paper down.

    But critics cannot simply say there might be a problem with a paper, they should make a clear case of where the paper has fallen down. They also have to make the author aware of the problem. This is often done privately, but if the authors fail to respond a critic can publish a rejoinder paper, again peer reviewed. It is then up to the authors to respond, correct or ignore.

    This is a process with clear parallels in the legal process. The burden of proof is on whoever is making a new claim, and the claims have to be openly published as a clear case. The initial filter of peer review in science is not that different from what lawyers will first assess when a client asks for representation- is there a coherent case in the first place as judged by experts in the field?

    By lowering the barriers to publishing any criticism of a scientific paper, blogs have created some quite severe problems. Firstly, the quality control has mostly gone so there are plenty of incoherent claims made. Secondly, the claims are not made as a clear case, but often as a scattered list of posts and comments. Thirdly, authors cannot be expected to respond to every criticism made on any blog.

    On top of these problems is the huge difficulty of producing from the blogs a reference database comparable to that of the published scientific literature.

    Blogs definitely have their place, but I don’t think they are quite ready to supplant traditional peer-reviewed publication as the best way to forward and share scientific knowledge.

  274. Tom, #21973.
    You say: “Not my area, hence I’ve no immediate response.” Well, I urge you to examine the matter. It is totally unacceptable to inflate claims of proxy significance by including instrumental information to gross up the results. Each proxy subset has very different properties.

    You ask: “Do you have any answer to the number of random proxies which would pass the full screening? I presume this is in your area of expertise.” Sorting out weird Mannian methods is quite time-consuming. In my opinion, the primary responsibility for providing accurate benchmarks lies with the authors and reviewers. We showed that Mann used a pick-two method and did not allow for this in his benchmarks. In order to check things, you’d have to re-do the entire calculation from scratch removing the pick-two calculation – something that I would have done if I could clone myself. But really – it’s outrageous for you to be criticizing me for not doing this calculation, when you have no words of criticism for the publication of false benchmarks.

    You observe: “You also haven’t responded to the the historical repeated divergence between younger and older trees in your combined Yamal-Khadyta chronology, despite your unsupported claim to the contrary at http://www.climateaudit.org/?p=7278.” I’ve been working on other things. My concern with your approach – as I mentioned to you – is that your effort to draw a distinction between old trees and young trees as indicators is not one that is drawn by dendros themselves. If your point were held out as being the “right” way of doing things, it would require re-doing all the other dendrochronologies in the world, something that I suspect that dendros would be reluctant to do, and a project that might yield unexpected results when applied to other data sets.

    I generally limit my sensitivities to techniques that are actually used by dendros. In the case at hand, Briffa had added a Schweingruber data set to Taimyr that was quite far away, while the Schweingruber Yamal site appears to be located between sites used in the Briffa Yamal data set. So my variation seemed to be within established practice, while yours is one that does not appear within the literature.

    The other larger issue that you’re overlooking was the original issue of the Yamal Substitution, which is related but different to the Schweingruber Variation, but one which was very much on my mind when D’Arrigo et al 2006 and Osborn and Briffa 2006 came out and in my submission of AR4 Review Comments.

    Having said that, some time in the next week or two, I’ll try to figure out what you did.

    http://www.climateaudit.org/?p=4539

  275. “The burden of proof is on whoever is making a new claim”

    What happened to the burden of proof on the original claim?

    Did it go *poof* just because someone else made a different claim?

    Magic?

    Andrew

  276. Tom P:

    “Once the paper is published, it is up to the wider scientific community to criticise the paper. Science is essentially adversarial – it doesn’t matter how many people a paper convinces, it can only take one critic with a clear refutation to bring a paper down.”

    And it does not make ANY difference where this refutation is published whether it on a blog, private emails or in the journal. The only thing that matters is whether it is more convincing than the original paper.

    Also SteveMC did go through the proper procedures in case of Mann 2008 but the process is rediculously inadequate (e.g. 250 word limit). This means people who want to understand the issue have no choice but to read Steve’s blog. If you have a problem with this then you should be taking it up with the editors of PNAS and other journals who clearly do not want people to point out errors in papers they accepted.

  278. Tom P wishes:
    “critics cannot simply say there might be a problem with a paper”
    .
    Replace “might” with “is” and you have a debate. Critics most certainly can, and do “simply say there is a problem with a paper.” And if the problem is fatal the paper needs to be retracted. Scientific knowledge grows by refutation as much as by conjecture. There is no obligation for a refutationist to re-conjecture.

  279. S. McIntyre says to Tom P:
    “My concern with your approach – as I mentioned to you – is that your effort to draw a distinction between old trees and young trees as indicators is not one that is drawn by dendros themselves.”
    .
    I can imagine Tom P taking exception, and perhaps we will crosspost. What McIntyre means. Tom P, is that although dendros know about the issue and some might care about it, chronologies are not always archived in a way that you can estimate or remove the effect. For you to require that such effects be accounted for would be a novel suggestion that would find quite a bit of resistance. Do you know how many chronologies there are in the ITRDB archive? Does it make sense for you to insist that McIntyre use a method that is not standard practice?

  280. Alan S. Blue (Comment#21989) says:

    “You can take bender’s statement one more step. The further step is: Treelines move.”
    .
    That treelines move proves the uniformitarian assumption does not hold. But this, too, is insufficiently black-and-white. The question is how *much* do treelines move, how *much* is the temperature response dulled or sharpened as treeline moves downslope or upslope? How bad an approximation is the uniformitarian model?
    .
    The issues are all *quantitative*. Leave your rheotric at home. This is about numerical analysis.

  281. Steve McIntyre (Comment#21993)

    “In order to check things, you’d have to re-do the entire calculation from scratch removing the pick-two calculation – something that I would have done if I could clone myself. But really – it’s outrageous for you to be criticizing me for not doing this calculation, when you have no words of criticism for the publication of false benchmarks.”

    Outrageous? I’m not willing to prejudge your results.

    Do you think you might get a little further along if you concentrated on completing your criticism of one paper before moving on to the next? Your answer does rather depend on whether you’d like to look back on a set of important publications in this area or some nice blog readership statistics.

  282. Tom P is getting cockier. He’s got more of the RC snark happening now. Keep it up, Tom P. Move off the facts and onto your fictions. This is surely the best way to help the alarmist agenda. Good work.

  283. Tom P asks for clarification:
    “Outrageous?”
    .
    Yes, it is baldly “outrageous” for you to demand somebody fix someone else’s fatal error when you do not do not criticize the error itself. (Never mind that the error can’t actually be fixed when the code is not available!).
    .
    [Pause 10 sec.] Yep. That’s definitely “outrageous”. The hypocritical height of all double-standards. That one goes in the database.

  284. Tom P asks:
    “Do you think you might get a little further along if you concentrated on completing your criticism of one paper before moving on to the next?”
    .
    How ironic. I was wondering the same of you.

  285. Tom, I view the original authors as having the primary obligation to do things right.

    Mann’s inclusion of Luterbacher series in the supposed significance result obviously tricked you. The responsibility for that lies with Mann, not with me.

    Mann’s use of pick two correlation in “observed” and pick one in benchmark also tricked you. It’s his responsibility to do it right.

    The obligation of original authors to do things properly is entirely independent of whether I submit something at a journal, write it up at a blog or do nothing. Your continuing failure to recognize the obligation of the original authors is disappointing.

  286. Tom Curtis–
    Lucia is a female name, and I am a woman. 🙂

    It is traditional to test statistical methods against noise. The rest of what you say reveals a profound lack of understanding of significance testing. It is traditional to test what a statistical method against “noise”. Even if trees do contain some signal, if that signal cannot be distinguished from noise, then you can’t conclude anything. Basically, if the same result would be obtained from noise and trees, then when you get the result, you must say, “This result may be ‘signal’ or it may just be noise. I don’t know which.”

    That’s call lack of statistical significance.

  287. S. McIntyre says:

    “Mann’s inclusion of Luterbacher series in the supposed significance result obviously tricked you.”

    “Mann’s use of pick two correlation in “observed” and pick one in benchmark also tricked you.”

    Tom P also wants to know if these tricks “matter”. But he thinks it’s McIntyre’s job to prove they matter. But that’s not the way it works. The burden of proof lies on the authors to show that these tricks do *not* matter. Moreover, if they had lived up to their obligation to make their methods transparent, anyone could check in an instant whether or not they matter. That we can’t check points to a failure to adequately describe methods. Thus there are at least three reasons why this paper never should have been accepted.

  288. TomP

    Blogs definitely have their place, but I don’t think they are quite ready to supplant traditional peer-reviewed publication as the best way to forward and share scientific knowledge.

    Talk about a strawman! who says blogs every will be ready to supplant peer-reviewed publications? Blogs are carving out their own role. You don’t like it; other people do. But blogs will stay here, so put on your big boy pants and deal with it.

    You also greately overestimate what has been “proven” when a paper passes peer review. The hurdle is that the paper had proven to an editor that it is interesting and useful enough to publish. No journal vouches that all their papers will be correct. Neither do the peer reviewers.

    It is later readers who discuss the paper etc. who ultimately decide if they are convinced. They get to make their own judgement. Period. Any other claims are silly.

  289. So Tom you’re saying that pick 2 which clearly modifies the thresholds presented in Mann08 and was not reported in Mann08 is not a red flag??

    How about potentially problematic?

    Possibly perplexing?

    Proximally imprudent?

  290. bender–
    Papers are almost never retracted. Even massively wrong papers are rarely retracted. (I think papers get retracted in Mathematics, where right and wrong can be obvious. But otherwise, it is very, very rare.)

  291. TomP,

    “Do you think you might get a little further along if you concentrated on completing your criticism of one paper before moving on to the next?”

    You’ve failed to support poor, if not actually fraudulent, papers with your arm waving.

    Now you pretend to become Steve’s Life Coach????

  292. “Blogs definitely have their place, but I don’t think they are quite ready to supplant traditional peer-reviewed publication as the best way to forward and share scientific knowledge.”

    Scientific truth is not a respecter of human institutions. Sure there are hundreds of thousands of spurious and ridiculous claims on the interwebs, but there is an unmatched diversity of viewpoint. I don’t blame news organizations, and govts, for example, who depend on peer reviewed science, but the scientists themselves ought to be able to fight their battles on the ‘net. And if they are in the right, these battles should be easily won. The hockey stick is an important rhetorical graph. It should be publicly defensible. Trillions of dollars in spending and taxes depend on it. These kinds of decisions can’t be left to a priesthood, which history has shown leads to corruption, time after time. Arguments that can be presented in lay terms should be able to be refuted in lay terms. No such luck with the hokey stick.

  293. lucia,
    Yes, retractions don’t happen in the soft/earth sciences. All the more reason to insist on (1) full disclosure of methods, and (2) sensitivity testing, as part of the standard burden of proof on the proponents of a paper. To claim that errors x,y,z “don’t matter”, this requires proof. Reviewers that don’t ask for it, authors that won’t provide it, Editors that don’t enforce it – well, there is a place large enough to hold all of them.

  294. And note that my view is the consensus view. Why else did Mann provide his (poorly conceived) “sensitivity test” in the Fig S8? Either he knew to do it, or a reviewer told him to do it. Either way, I’m right.
    .
    Paying lip service to sensitivity testing is increasingly required to get a paper published.

  295. TomP,

    “…But really – it’s outrageous for you to be criticizing me for not doing this calculation, when you have no words of criticism for the publication of false benchmarks.”
    Outrageous? I’m not willing to prejudge your results.”

    Typical DISINGENOUS response. Absolutely IGNORE ANYTHING THAT YOU DO NOT WISH TO DISCUSS AND WHICH DESTROYS THE BASIS FOR YOUR ARGUMENTS!!

    Has anyone else called you a nitpicker recently??

    You focus on one specific issue that Steve has not looked into while ignoring or apologising for MAJOR flaws in the same papers.

    If you were a defense attorney you would be contesting her parking tickets in court instead of the Premeditated Murder charge for which she was actually arraigned.

    If you have an issue with the young/old tree growth patterns, shouldn’t you be belaboring the Dendro community who are the “experts”?

    Should you not also be belaboring the Hockey Stick authors who have used the tree cores with no efforts to take your issue into account??

    Instead, you come here whining about the one person who HAS spent a huge amount of personal time in trying to clean up some serious problems in the Dendro communities published work.

    What is you excuse for this???

  296. What Tom P is doing is very simple. He is trying to shift the burden of proof from Mann onto McIntyre. If there were no readily identifiable flaws in Mann’s paper, this would be reasonable (and bugs’s admonition to sweat a little would be in order). But the flaws have been identified and they are disconcerting. Their impacts have not been calculated. But they should be. And Mann should be as concerned as anyone over accuracy. Does he even care that these errors are in his paper? Is he so self-confident that he thinks no amount of error will “matter”? Look, Mike, if it’s a pre-ordained conclusion, then why are you bothering with the whole data thingy?

  297. Aren’t published papers just superseded not retracted? Papers can be useful for a while and, as such, are recognized by citations. After a period, other more useful papers are written which supplant even the most successful paper.

    There is a fundamental issue here. Papers are not ‘correct’ or ‘incorrect’. They do not contain ‘truth’. Every paper that has been published in history has been or will be superseded by others. There is no burden of proof since there is no issue of ‘truth’. Scientific papers are either useful or they are not useful in the continuing effort to create clarity around an issue. As a peer reviewer, I have recommended papers that I do not agree with for publication. I did this not because they contained the ‘truth’ (I personally disagreed with their conclusions) but because their publication would be useful to the ongoing work in my field.

    Papers that do not allow for replication cannot be considered useful. They do not provide to the continuing effort.

  298. “Papers are not ‘correct’ or ‘incorrect’. They do not contain ‘truth’.”

    This is a far cry from saying that ‘truth’ does not exist, and that the goal of science is not to asymptotically approach that ‘truth.’ Papers are helpful or not based on how they serve that goal.

  299. Blogs definitely have their place, but I don’t think they are quite ready to supplant traditional peer-reviewed publication as the best way to forward and share scientific knowledge

    Peer-reviewed publications has never been the best way to forward and share scientific knowledge. The book ‘Proofs and Refutations” by the noted philosopher of science Imre Lakatos The book traces the historical development of mathematical knowledge and shows how this develops as interactions in informal (i.e. outside of the published journal articles) mathematical research. Learned articles are published which are perfectly valid as they stand. Other researcher notice inadequacies and modify concepts to create deeper and more powerful ideas. This work is not done in peer-reviewed learned publications since the published proofs are valid as they stand but are not useful for further research.

    Researchers hold conferences, workshops etc to foster this informal interaction. They do not spend their days locked in their office poring over the latest journal articles. Most useful work at conferences is not done in the technical sessions but in the corridor chatter between them. Blogs are not replacing the learned journals but are supplementing the hallways outside of conference rooms where most of the useful interactions in research go on.

  300. There’s another aspect to the pick-two thing. Does anyone, including Tom P, seriously think about why on earth a pick-two comparison was used in the first place? Does anyone think that this was the original methodological design – that, instead of calculating the correlation to the gridcell, they’d pick the highest correlation from 2 neighboring gridcells (with the choice of nearest neighbor sometimes turning on quirky Matlab rounding in the 14th decimal place)? What rationale would there possibly be for introducing such a method? Does anyone think that, just maybe, before using the pick-two “method”, that Mann might have first done the obvious comparison to the actual gridcell?

    Does anyone think that it might be remotely possible that the passing yield from code 9000 dendros in the obvious calculation was calculated and discovered not to be significant and that the yield was “improved” using the pick-two method?

    Tom P, if Mann calculated the results for dendro proxies using the “natural” gridcell correlation, do you think that he would have had an obligation to report this? Alternatively, if you think that he didn’t do such a calculation, do you think that it was negligent practice not to do so?

  301. TAG,

    What you are saying is that some important content of any particluar published paper may be subjective.

    That’s a problem.

    Hence the inherent logical fallacy in Appealing to Published Papers themselves, not the facts (truth) they may (hopefully) contain.

    Andrew

  302. With regard to the burden of proof, before an expert is allowed to testify in a trial, the expert must disclose all of the data and calculations upon which his opinion is based under Federal Rule of Evidence 705. All States have a similar rule. Here is an example from a real case:

    “For two reasons we must reverse the judgment and remand the case for a new trial. The first reason is that the person the county called as a witness as to the value of the property was unable or unwilling to provide the court and jury with proper documentation and support for his opinion as to the value of the property being taken. Although he said that he ‘used market data, interviews with [persons] and other materials . . . to come up with these adjustments,’ on cross-examination he would not or could not produce any of the market data or names of persons to support his statements. His testimony was inherently incredible.”

    John Law Ayers Estate v. Hernando County, 706 So.2d 349 (Fla. 5th DCA 1998). In other words, the trial court should have excluded the expert’s testimony because he failed to provide the data and calculations underlying his opinion. Why should climate scientists be afforded credibility by the fact of mere publication without disclosing their data and calculations? An opinion lacking the support of data and calculations is, as the court stated, “inherently incredible”.

    Regards,

    WJR

  303. TAG
    Retractions sometimes (albeit rarely) DO happen. When someone’s methods are SOOO bad that there is nothing trustworthy in the data. This paper is not that bad. But there are enough questionable elements that follow-up is required. People are calling for McIntyre to write the supersedent. That’s like asking the IRS auditor to adjust your tax return. Or the mailman to send you a Christmas present.

  304. Steve McIntyre asks:
    “why pick-2?”

    It’s obvious why someone would use a pick-two rule. They tried pick-one and it didn’t give “the right signal”. Do you think pick-three would work even better?
    .
    I’m sure there are “justifications” for the rule. The problem is the benchmarks need to move with the method.
    .
    I can kick a field goal from mid-field … if you set the posts on the 40-yard line.

  305. Will J. Richardon,
    So if the lawerly “burden of proof” were to apply to science, then a failure to meet a burden of proof within a scientific paper would not excuse that burden. The testimony would be considered “incredible” until such proof was forthcoming.
    .
    Chew on that, Tom P.
    .
    The case of the truly incredible vanishing & re-appearing hockey stick.

  306. Is it possible that Mann actually *tricked himself* into cherry-picking, by not realizing that the benchmarks were moving with the pick-two rule?

  307. With regard to the continuing dissection of ‘proof’ and by implication ‘truth’, one can pose the question of just why is this research being funded. Does anyone reading this really care about the latest application of statistics to dendro-climatalogical result? Does anyone really care what the mean world temperature was in the year 1453? People care about these results because they are being used to inform public policy. The criterion that should be used to judge them is if they are useful to inform economic policy related to the response to the AGW concern.

    The question is not if they are ‘true’ or not but whether they are good enough for public policy. Without replication, how are we to know? If they use a novel statistical technique which is unknown to statistical authorities the how are we to know? If data has been eliminated by obscure techniques then how are we to know?

    Papers from this research effort should be evaluated against the criterion as to their utility in the shaping of public policy. Their place in the search for the ultimate scientific ‘truth’ (assuming that such a concept has any meaning at all) is of collateral interest only.

  308. TAG, that is how the research is being evaluated: is the data unambiguous enough to conclude with certainty that current temperatures are “unprecedented”? There’s no doubt they’re high: we’re in an interglacial period. But are they “unprecedented” – or “alarmingly higher than high”.

  309. S. McIntyre asks:
    “if Mann calculated the results for dendro proxies using the “natural” gridcell correlation, do you think that he would have had an obligation to report this? Alternatively, if you think that he didn’t do such a calculation, do you think that it was negligent practice not to do so?”

    Mann should have done the calculation and should have reported it. The pick-two method is “bizarre” and the failure to report the pick-one result is negligence, on the part of the authors, reviewers, and Editor.

  310. “Does anyone really care what the mean world temperature was in the year 1453? ”

    “He who controls the past, controls the future, he who controls the present controls the past. ” Orwell, 1984

    This fight is about control, nothing less. We have a priesthood seeking to impose a finding, with huge financial and geopolitical ramifications on a the democratic west without open debate. I don’t know why Steve wastes his time arguing with “Tom P”, unless he knows who he is and is trying to get him on record to say indefensible things. Tom P cares about the ‘revolutionary truth’ of AGW more than he does the bourgeois truth of everyday facts and evidence. That’s why he will not acknowledge being caught out. I don’t see why Tom P would ever find it in his interest to concede any point. To quote another line from 1984 “Freedom is the right to say 2+2=4”. when Winston Smith is finally crushed by the system, he writes 2+2=5, and does not question it. We skeptics are all Winston Smiths, and we know what they plan for us.

  311. Steve McIntyre (Comment#22004)

    “The obligation of original authors to do things properly is entirely independent of whether I submit something at a journal, write it up at a blog or do nothing.”

    I completely agree. But errors for whatever reason do occur and then the question is how best to identify and resolve these errors. I certainly have issues with some of the errors you think you have identified, but, I’m equally concerned about the very way you are going about this.

    Having to issue multiple clarifications to the media when they misinterpret your results should indicate that there might be a better way of proceeding.

  312. ““Does anyone really care what the mean world temperature was in the year 1453? ”

    “He who controls the past, controls the future, he who controls the present controls the past. ” Orwell, 1984

    This fight is about control, nothing less.”
    .
    Can we focus on the fight, not the war? This fight is about data. Over what time scale are current temperatures “unprecedented”? 400 years? 600 years? 1000 years? 2000 years? 10000 years? How does the number coming out of the analysis depend on what goes in? Remember: GIGO.

  313. Tom P whines:
    “Having to issue multiple clarifications to the media when they misinterpret your results should indicate that there might be a better way of proceeding. ”
    .
    Back to Briffa now? McIntyre followed the rules and got over-ruled. Or did you not read the story about the IPCC lead author that was editing with a heavy hand reviews of his own work? The clarification that the media *really* needs is still to come. You harp on about the “false” claim of the broken hockey stick if you like. Not only is it false – and not only will this be shown in due time – but the story gets ugly.

  314. Tom P:
    There are two false hockey stick graphs in the Yamal thread at realclimate where the graphics need to be amended. Could you look into that? They are the ones captioned Mann et al. (2008) and Kaufman et al. (2009).

  315. Steve McIntyre (Comment#22019) October 19th, 2009 at 10:18 am

    Now you’re letting it out. It’s starting to sound like me.

    In my opinion it is a self proving shenanigan. If we assume Mann had the complete integrity befitting a scientist of his stature and was simply incompetent in the realization that picking 2 gridcells would effectively bias the results towards more proxies passing then why would he spend all the extra time to write the code to pick two separate gridcells.

    Of course if we change the primary assumption the whole thing makes sense.

    Looks like a duck to me. Now my question is why would a group feel such pressure that they need to do something like that? Certainly if they did it, they should simply report it and the effects. What’s the big deal if the past is slightly warmer than present, can’t they still rely on the cause and effect for AGW? The ramp rate of AGW and the future projections from computer models that make hockey sticks look like golf clubs.

  316. Carrick (Comment#21980) October 19th, 2009 at 7:25 am

    You will note that I commented upon this aspect of Tom Ps behavior over at CA. It’s a denialist tactic. You can visit the threads here and find similar behavior on the 2nd law threads. Eventually, Tom will change the subject or leave the thread. The topic is already morphing a bit ( note how we talk about Tom behavior and what it will take for him to admit he is wrong ) Trust me for a variety of reason’s you will never get Tom to say the simple words: I (tom) was wrong or Mann was wrong. Never. In short there is NOTHING will cause him to change his beliefs. That means his beliefs, as far as he is concerned, are not falsifiable. He has faith. Of course, I could be wrong about this. We will see.

  317. Jeff Id (Comment#22034) October 19th, 2009 at 11:59 am

    I can think of one rationale for doing pick 2 that would almost make sense. You write your code and you find that one gridcell is really close to a border maybe on a border. And that gets you to thinking… gridding is arbitary.. so how do pick which grid when a site is close to a boundary? In any case you would have to report the change in stats from doing a multipick.. But in mann mind ( and Tom Ps ) there is no problem with this pick two strategy. it’s just an extension of picking cores that correlate.

    Anyways, I would suppose one “right” way to do things would be to regrid the temperature series or something like that.. having the hadcru code would make that easier.

  318. steven mosher (Comment#22036)

    mosh, I agree it may be justifiable. But then you”ve got to move the benchmarks the right distance by tracking how many times you accept the second pick. They didn’t do that. It is possible they honestly didn’t realize they were biasing the stats. Stuff like that happens when papers are hastily prepared.

  319. TomP,

    “Having to issue multiple clarifications to the media when they misinterpret your results should indicate that there might be a better way of proceeding.”

    But, Mann and Briffa have never issued clarifications. They continue to claim infallibility.

    Again, you continue to nit-pick the auditor and not the guilty being audited. I would ask you exactly what you are trying to accomplish??

  320. kuhnkat: Tom P is vanely trying to defend Briffa and Mann as best he can. What he doesn’t understand is that their errors of process, which are indisputible, are indefensible. His misguided efforts are doing them more harm than good because of his inexpert approach. The best possible outcome for the Team would be to recognize the errors, consider them in a careful re-analysis, and hope for an unchanged result (ain’t gonna happen). Tom P’s special pleading does them more harm than good. But he doesn’t see it that way.
    .
    mosh: the easiest justification for pick-n is “teleconnection”.

  321. TomP=Propagandist/Distracter

    Does he believe what he posts? Prolly not.

    The result of his comments may be that he keeps reminding people of how unreasonable AGWarmers are, and reinforcing what a joke AGW is.

    Andrew

  322. Bender,

    “Tom P wishes:
    “critics cannot simply say there might be a problem with a paper”
    .
    Replace “might” with “is” and you have a debate.”

    critics cannot simply say there is be a problem with a paper

    (sorry, I am an old smart aleck without the smart)

  323. kuhnkat (Comment#22042)
    Thanks, the point of correction is justified. Critics can’t “simply say there might be a problem with a paper” and expect themselves to be believed. But it’s more of less obvious that hand-waving is not a proof. I don’t think we’re here to debate that, are we? McIntyre is not just hand-waving.
    .
    If OTOH Tom P is arguing that only published refutations can find their way into the IPCC process, he’s kinda right, but not exactly.

  325. bender (Comment#22037) October 19th, 2009 at 12:38 pm

    I can see it happening ( I did something like once) innocently, but once the ramifications are pointed out a correction is in order. In my case It was easy because I just changed proceedures for subsequent tests and pointed out the error in my first set of results and reran my first set of results to show the effect of post hoc screening choices. It’s a hazard of doing exploratory work or working on the edge of science where the cookbook has no recipes or the only recipes it has are not giving you the results you expect and “know” to be true. hehe.

  327. Bender,

    Yes, it is quite interesting that TomP implies that Steve has only made “claims”of error when Steve has written up the issues in detail, and even gotten a couple published. In fact, Steve has done such a good job that many people, like me, find them easily, understandable.

    Then there is the fact that TomP makes a claim of error, throws up a couple of disconnected graphs and a bunch of arm waving and then expects Steve to put his life on hold until TomP is satisfied!!!

    I am not sure that double standard even begins to cover this hubris.

  328. Steve McIntyre (Comment#22019)

    “Tom P, if Mann calculated the results for dendro proxies using the “natural” gridcell correlation, do you think that he would have had an obligation to report this? Alternatively, if you think that he didn’t do such a calculation, do you think that it was negligent practice not to do so?”

    Unlike your previous criticism that compared included actual proxy to random proxy inclusion ratios without using equivalent screening, this is a valid point. Of course a valid point, such as the inhomogeneity in the Yamal core ages, does not mean there is a problem – there wasn’t one in the case of the Yamal chronologies.

    I would presume that screening to correlation of just the nearest grid point included too few proxies at the stated significance level (P < 0.1) and so this screening criterion was relaxed to include the next nearest point. I agree* it would be good to make the reason for the relaxation explicit.

    This relaxation would increase the number of proxies that would pass the screening process by chance alone, and so should be taken into account in the final proportion of 13% given. Has there been a calculation to show that this proportion has been incorrectly derived given the relaxation in the screening?

    * But why use the word "negligent"? I'm able to state that you shouldn't have kept using the 13% figure after additional screening without using pejorative language.

  329. Tom P,
    get your aorry ass over to CA and read about the Yamal substitution and stop pissing around.

  330. Steve, bender, others:
    You are grossly underestimating the creativeness of Mann 😉 It is not simple pick-n -rule that has actually been coded, these are the relevant parts of the code (propick.m):

    igrid = 2; %numbers of gridpoints near proxy site used to cal. r
    igr = 1; % numbers of gripoints need to pass the CL for selecting a proxy series
    —
    if sum(d3)>=igr % at least ‘igr’ correlation of ‘igrid’ pass the CL

    In other words, the code allows for any pick-n-out-of-m -selection rule. Who knows what combinations have been tried before settling to pick-two rule 😉

  331. I think that I shall never see
    A T-stick lousy as a tree.
    For trees will thrive in damp or heat,
    Adapting to the climes they meet.
    And bears go in the woods, they say,
    To fertilize them night and day.
    Each bruin finds the perfect spot,
    In weather cold or weather hot.

    Pimply PhD’s in sandals
    Search’d for temper-ature handles,
    Looking for some sort of proxy,
    Whether good or whether poxy:
    Maybe how lake silt behaves?
    Or bat poop laminae in caves?
    At last, they thought they’d found the key
    in Dildoclimatology.

  332. Tom P asks:
    “Has there been a calculation to show that this proportion has been incorrectly derived given the relaxation in the screening?”
    .
    Gee Tom, I dunno. Does Mann et al show this calculation in their paper? Woulda been nice, hunh?

  333. When it came time to correlate ring size to temperatures, it would seem an awful lot of trees didn’t get the memo.

  334. Does anyone know who TomP really is? Just some guy who wanders around climate blogs arguing idiocy? Does he have a job? If so what does he do? Does he have any technical background? Has he ever actually done science or engineering? Or is he trained only in one of the (very) soft sciences like environmental studies or psychology? Is TomP just the pen name of some post doc who works for Gavin, or (worse) Mann?

    Come on TomP, who the heck are you? Since you so very much like to point to “authority” as proof of credibility, show some cred man.

  335. Tom P is a loose collection of incoherent statements, graphs, and links that Gavin Schmidt cuddled up to – until he realized Tom P was toxic. His “credentials” are that he is our friend. Perhaps he is the incompetence that lurks in all of us. We may not like him. But he is part of us.

  336. Bender,

    “Perhaps he is the incompetence that lurks in all of us. We may not like him. But he is part of us.”

    Very philosophical, but I would still like to know who the heck TomP is, and more importantly, who the heck he thinks he is. I have never seen anyone on a blog waste so much of people’s valuable time. Step up to the plate man, take the snap, show some courage… who the heck do you think you are?

  337. “Can we focus on the fight, not the war?” – bender

    Good point. I just don’t know why you are continuing the fight with somebody who has no interest in conceding the truth. I think the fight is long over on the subject of potential selection bias. In my humble opinion, the fight now is over autocorrelation. This is just another example of assuming the result in the premise, so to speak. It is always dangerous to declare a negative, but Gavin did claim that the literature supported the idea that there was no red noise in tree ring data outside of temperature biases. I think that this is where the fight is. Once autocorrelation is gone, the whole make a hockey stick out of anything argument disappears. The fight is not over ’til this issue is dealt with.

  338. Very philosophical, but I would still like to know who the heck TomP is, and more importantly, who the heck he thinks he is. I have never seen anyone on a blog waste so much of people’s valuable time. Step up to the plate man, take the snap, show some courage… who the heck do you think you are?

    WTF?

  339. “Gavin did claim that the literature supported the idea that there was no red noise in tree ring data outside of temperature biases.”
    .
    Really? Where? Better ask Gavin to go talk to Hal Fritts, who pioneeered the idea of “physiological preconditioning” – aka red noise.

  340. “I just don’t know why you are continuing the fight with somebody who has no interest in conceding the truth. ”
    .
    Refuting the imbeciles with sensible examples may not help the imbeciles, but it does help everyone else. And it’s not ALL about red noise. There IS some signal in there. That’s the problem. If all you want to talk about is red noise you are going to get no traction with real dendroclimatologists, who know for a fact that there IS some signal in there, however weak.

  342. Didn’t Lucia just explain to you Boris? Do you not understand English?

    And BTW, I read on all kinds of AGW blogs this argument that skeptics don’t publish. However, I’ve read many skeptic papers in publication and I’m aware of 2 reviews of Steve Mc’s papers receiving what are biased unfair reviews designed to prevent publication. It turns out you have to have a stick with a blade to talk about hockey. Even if your point is that the stick is bent.

    I don’t buy it. If someone conspired and gave me a bad review, I’d make damn sure I published that paper–somewhere. I’d even hide some dirty acrostics in it too.

    I’m not saying Mann is right because he got his paper published. I’m not saying anyone is wrong for being unpublished. But the constant chorus of “fatal flaws” rings mighty frickin hollow when people aren’t even writing up their arguments in a serious way so that star dendro kid from India can know there’s even a problem when he checks Web of Science.

    People see all the political discussion on blogs and they ignore it. The signal, if there is one, is too hard to find in the noise.

  343. Step up to the plate man, take the snap, show some courage… who the heck do you think you are?

    Who is bender?
    Who is TCO?
    Who is Boris?
    Who is Lucy Skywalker?

  344. Boris:

    People see all the political discussion on blogs and they ignore it

    Feel free to ignore this blog then. 😛

    Seriously though most science gets done away from the peer review process. Peer review is where you generally archive your work for antiquity, not where you break new ground—or at least not often, I’ve had a few occasions where this has happened with my published papers creating a bit of a stir, but generally people have heard your work ad nauseam at conferences before it ever gets published.

    I have occasionally published refutations of other peoples work, but it’s not a pleasant thing to do, and in some instances created a bit of hard feelings. More frequently papers get demolished at conferences, or obliquely in a series of other publications that point-by-point refute the findings of badly performed research.

    I trust the scientific process in this case, if there is something to what people are saying, to take care of this for us. Inevitably if nobody here publishes this, and there is substance to the criticism, somebody else will publish it and claim priority. (That latter has even happened to me a few times where I talked about my work before I published it.)

  345. Boris,

    “Who is bender?
    Who is TCO?
    Who is Boris?
    Who is Lucy Skywalker?”

    Some are people willing to engage in constructive discussion, some only want to add noise.

    “People see all the political discussion on blogs and they ignore it. The signal, if there is one, is too hard to find in the noise.”

    If you think that is true, then why are you wasting time commenting here (or at any other blog for that matter), especially since it appears that you add mostly noise to the thread?

  346. It is always dangerous to declare a negative, but Gavin did claim that the literature supported the idea that there was no red noise in tree ring data outside of temperature biases. I think that this is where the fight is. Once autocorrelation is gone, the whole make a hockey stick out of anything argument disappears. The fight is not over ’til this issue is dealt with.

    .
    The “no red noise” argument is just a slick way of diverting the argument. It doesn’t matter if there is no true “red noise” (where “red noise” is defined as autocorrelation of the residuals once all climatic factors have been accounted for) in tree rings. First, determining whether this is true is not possible until one actually measures the residuals once all climatic factors have been accounted for. This has never been done and likely will never be done.
    .
    Second, it doesn’t matter. All that matters is that there is at least some signal with autocorrelation that is not the signal of interest. Obvious culprits are precipitation and CO2. CO2 definitely shows autocorrelation, and precipitation often does as well (see Cook’s drought reconstructions or the work of Koutsoyannis). Unless those signals are known and quantified in a particular tree ring series, they are, for all practical purposes, red noise.
    .
    So while – given the definition of “red noise” above – there may not be any in tree rings, the unknown “signals” in tree rings are autocorrelated and are unknown and therefore may be modeled by . . .
    .
    . . . hold on to your hats . . .
    .
    . . . RED NOISE!

  347. “who know for a fact that there IS some signal in there, however weak.”

    Sure. I don’t doubt it for a second. What I do doubt, and what Mann and Gavin seem to claim, is that it is the *only* autocorrelation. Or at least the strongest. If that is correct, then Steve’s Argument, while an interesting cautionary tale with regard the method employed, has no bearing on this case. This isn’t about winning the argument, it is about getting closer to a correct interpretation of the data.

  348. All that matters is that there is at least some signal with autocorrelation that is not the signal of interest.

    True enough, I can think of plenty of them myself. I just can’t prove any of them, or disprove Gavin and Mann’s contention. Harping on a settled argument about selection bias isn’t getting any closer to ending the controversy once and for all. If the vast majority of non climate noise is white, the stick stands. This cute little demonstration or no.

  349. >If you want a code to produce that result I suggest your time might be more healthily spent learning R rather than indulging in this obsessive behaviour.

    Tom, did you ever fix your R code as I requested? You have to change your Yamal ‘robustness’ test to properly account for the age of the trees.

    And why do you complain that Steve jumps from subject to subject, wen you do the same thing?

    If you are jumping to another blog, I recommend chiefio.

  350. If the vast majority of non climate noise is white, the stick stands.

    .
    Eh?
    .
    CO2 ain’t white. Precipitation ain’t white. Neither of those have been separated from the “signal” (read “temperature”). CO2 is quite definitely stick-like to begin with.
    .
    It’s not the non-climatic noise Mann should be worried about – despite what Gavin thinks he read.

  351. Ryan O:

    Second, it doesn’t matter. All that matters is that there is at least some signal with autocorrelation that is not the signal of interest. Obvious culprits are precipitation and CO2. CO2 definitely shows autocorrelation, and precipitation often does as well (see Cook’s drought reconstructions or the work of Koutsoyannis). Unless those signals are known and quantified in a particular tree ring series, they are, for all practical purposes, red noise.

    I should point out that if what you are trying to correlate to is global land+ocean mean temperature, then even local temperature variations can be a signal “not of interest”, since local temperature doesn’t always correlate that well with global temperature (especially over 1000-year spans).

    And I think the noise here is “pink” rather than red (1/f rather than 1/f^2). That certainly is the case for climate temperature series (at least the ones I’ve analyzed), I can’t say what happens for precipitation.

    You are right that the “there is no such thing as red noise” argument is (heh) just a red herring.

    It is well understood that these power laws “end” at sufficiently low frequency, in the so-called “turbulence source region” (where the power spectral density is more or less constant) and for frequencies below the source region, the PSD → 0.

  352. TJA,

    “Harping on a settled argument about selection bias isn’t getting any closer to ending the controversy once and for all.”

    Please explain why you think the argument about selection bias is settled.

  353. who know for a fact that there IS some signal in there, however weak

    For the record, I don’t think there is a signal in tree ring samples corresponding to global land+ocean mean temperature. None.

  354. Carrick,
    .
    It’s astounding how absolutely lame so many of the arguments are that justify what dendros do.

  355. Ryan O, I agree.

    The explanations that people give for expecting a correlation between global mean temperature and tree ring samples resembles something you’d expect to find in a very poorly written science fiction movie script. It’s that bad.

  356. You guys and gals are going to be playing whack-a-mole for some time to come. Not dendrochronology, but I am sure you’ll find these are wrong too.

    http://rintintin.colorado.edu/~axford/axford_et_al_QR_2009_CF8.pdf

    http://faculty.eas.ualberta.ca/wolfe/eprints/Briner_QR06.pdf

    I think the problem is that you need to engage in the debate in the venue that the debate actually takes place in. You can whine and moan about how awful dendrochronology is, but not many scientists will listen until you actually start publishing it. If it is so clearly wrong , why not take the time to write the paper and publish it.

  357. Carrick

    “Seriously though most science gets done away from the peer review process. Peer review is where you generally archive your work for antiquity, not where you break new ground—or at least not often, I’ve had a few occasions where this has happened with my published papers creating a bit of a stir, but generally people have heard your work ad nauseam at conferences before it ever gets published.”

    What a strange thing to say. Of course science gets ‘done’ away from Peer Review, you do the science first then submit it.
    As to ‘archiving your work for antiquity’ – well that’s a very strange thing to suggest. How do you expect people from other countries to know what you’re doing? Does everyone have to attend your conferences? It’s such a bizarre thing to say.

  358. You say:

    most science gets done away from the peer review process. Peer review is where you generally archive your work for antiquity, not where you break new ground—or at least not often, I’ve had a few occasions where this has happened with my published papers creating a bit of a stir, but generally people have heard your work ad nauseam at conferences before it ever gets published.

    Ross says that this is what happens in economics. I don’t think that that is what is happening in climate science right now. Nature and Science demand things that are “new”. As a result, we’re seeing Team reconstructions often with pretty hairy statistical methods being introduced in very senior publications without much, if any, prior discussion, let alone discussion ad nauseam.

    Given this situation, the analytic blogs are providing the after-publication discussion that really ought to have taken place previously.

    You mention conferences. However, IMO there is far more substantive exchange at the analytical blogs right now than in the conferences that I’ve been to (AGU). The analytical blogs are like an ongoing 24/7 seminar or online workshop. The unfortunate limitation right now, of course, is that, as a result of the policies of the opinion leaders in the field, mainstream climate scientists boycott these highly visible online discussions or, when they occasionally break the boycott, do little more than moon the discussions with adolescent drive-by comments, a spectacle that is usually perceived as exposing their shortcomings rather than the presumably hoped-for effect of generating shock and awe.

  359. Here is a quote from Esper, Cook, Krusic, Peters, and Schweingruber, “Tests of the RCS Method for Preserving Low Frequency Variability in Long Tree Ring Chronologies; Tree Ring Research; Vol. 59(2), 2003:

    “Every tree ring series, whether it is based on TRW or MXD, contains some non-climatic variations. This “noise” can be caused by site related effects (e.g. competition and disturbance) or biological effects (e.g. aging). As a consequence of integrating biological and ecological impacts within the measured parameter, TRW or MXD, variability in the resulting chronology will often depart from variability associated with climate. This departure from climate is often systematic and persistent over time.”

    –
    Sounds like an acknowledgement of the existence of red noise to me.

  360. Nathan:

    How do you expect people from other countries to know what you’re doing? Does everyone have to attend your conferences?

    In my area(s), yes, we do all attend the same conferences.

    It’s such a bizarre thing to say.

    Your opinion, so BFD on that.

  361. Layman Lurker (Comment#22083)
    October 19th, 2009 at 8:35 pm
    Not to say I pre-empted you, but Hal Fritts wrote the book on why trees, as biological integrators, are their *own source* of red noise. They take the white and pinkish environmental noise around them and very strognly redden it. No need to hunt for an external source.

  362. MikeN (Comment#22073) October 19th, 2009 at 6:45 pm

    ya Tom P is doing the blog bouncing think. After a while people will use a Tom P filter, like the TCO filter, you read the name of the poster and just wave and drive by. Sad, he had such potential. Kind of a less talented JohnV.

    why would you send him to chiefio.
    Smith will eat him alive. Tom doesnt have the attention to detail required to keep up with smith. It would be a massacre. Maybe Tom will end up over at Open Mind and he will post where the host will protect him from serious challenges.

  363. Boris (Comment#22067) October 19th, 2009 at 6:16 pm

    bender is my sockpuppet. duh. he even said so over at CA.

  365. Carrick

    Seriously? everyone in your field attends the same conferences? That’s a rather unique situation.

    The BFD is that most science types can’t attend every conference – often because they live in different countries – so your ‘publishing for archival purposes’ line may apply to you specifically but hardly applies universally.

  366. bender (Comment#22085)
    –
    Understood. BTW, I appreciate not only your knowledge, but your ability to argue and quickly paint the Tom P.’s of the world into logical corners. I would never presume to be able to scoop you on much – except maybe on how to use google. 😉

  367. lucia (Comment#22088)
    October 19th, 2009 at 9:07 pm
    Never mind that. We are teleconnected. I’m his poolboy.

  368. Steven McIntyre

    Oh come on buddy:
    “The unfortunate limitation right now, of course, is that, as a result of the policies of the opinion leaders in the field, mainstream climate scientists boycott these highly visible online discussions or, when they occasionally break the boycott, do little more than moon the discussions with adolescent drive-by comments, a spectacle that is usually perceived as exposing their shortcomings rather than the presumably hoped-for effect of generating shock and awe.”
    Ha ha haaaaaaa sounds like whiney whiney to me.
    Get off your butt and publish some results. Stop being a slacker and moaning how no one will discuss it at your house.

  369. bender (Comment#22063) October 19th, 2009 at 5:49 pm.

    err duh, isnt the whole point of a tree/forest to redden the enviromental signals. or am I being stupid again?

  372. Nathan:

    Seriously? everyone in your field attends the same conferences? That’s a rather unique situation.

    Over 25 years, I’ve done research in a wide range of areas, and it’s been my experience it works this way in practice in all of those.

    Nothing unique about it, any more than there is anything unique about your smug but ignorant arrogance.

  373. I don’t see a Nathan there.

    Surely as wise a one as you is published, yes?

    Apples to oranges comparison in any case.

    These guys could use some training in frequency analysis.

  374. Steve McIntyre:

    However, IMO there is far more substantive exchange at the analytical blogs right now than in the conferences that I’ve been to (AGU).

    The “meat” in that conference happens in the poster sessions. At workshops, they often have an evening that’s dedicated to “pressing research issues”. These tend to be high on flaming, but a lot of good discussion happens as well.

    I don’t see the opportunity for in depth sorts of discussions as have been appearing recently in the analytical blogs (as you so nicely describe them), of course. The biggest disadvantage of blogs is that people can post from anonymity, and that tends to encourage scurrilous behavior on the part of some.

  375. Carrick,
    I think you are the one generalising, and for whatever reason pretending that publishing is little more than ‘archiving’.

    No, I didn’t publish that paper (you may note that I said I hadn’t read it), but it’s certainly something you guys need to audit. Obviously they haven’t been reading your blog posts, and need to learn how to do science the real way.

    “Surely as wise a one as you is published, yes?”
    Ha ha! Look I never claimed to be wise, buthey you can think whatever you like.

  376. Carrick

    “Surely as wise a one as you is published, yes?”
    So I assume you have, so that’s good. So why don’t you publish on why Biffra is wrong?

  378. Nathan, I take that as a “no” as in you haven’t published.

    So seriously, you are here trying to lecture us on how science works, from a completely ignorant perspective?

  379. Nathan:

    So why don’t you publish on why Biffra is wrong?

    Why? Do you think Briffa is wrong?

  380. Carrick

    “So seriously, you are here trying to lecture us on how science works, from a completely ignorant perspective?”

    No, I am not ignorant of the process. I am not lecturing you on how science works. You made a claim that publishing was basically for archiving. I think we both know that that is not a sensible claim.

    “Why? Do you think Briffa is wrong?”
    I don’t know. It did seem from the Blah that was going on at CA that there was a ‘serious’ problem and that it was all wrong.
    Do you think it is wrong?

  381. Nathan, it is a sensible claim because it is true. And indeed you keep “lecturing” me on the process because you keep telling me what I know to be true, is false, all the while admitting no experience with the system!?

    But the fact is most scientists do go to at least one international conference each year, these typically happen 2-3 times a year. Most universities have grant monies set aside specifically for that, but I suppose you wouldn’t know that would you?

    Beyond that, you seem to be unaware of preprints, e.g., the arXig.org (what we used to call the XXX archive back when it was hosted at LANL). Again, these get widely circulated and discussed (as do, I might mention, the PowerPoint slides, which get disseminated to people who couldn’t make the conference), and by the time the research gets accepted and published, the contents of the paper are often well known within the community.

    Do you think it is wrong?

    You’d have to tell me which “it” you are referring to.

    I think it’s wrong to not release your data once you’ve published it. (Hardly necessary to put that simple observation into a peer-reviewed publication though.)

  382. What a great thread Lucia. It’s been one of my favorites in blog land.

    Here is another tweak to the story. I’ve got a plot of what happens to the signal as ever more red noise is added to it in CPS. Now that so many have considered the effects of various levels of noise and sorting check this out:

    http://noconsensus.wordpress.com/2009/08/26/more-on-cps-signal-distortion/

    And followed by a more subtle post:

    http://noconsensus.wordpress.com/2009/08/27/more-hockey-mathmagic/

    We all get so caught up in the postulation of possibilities (had to write it) it becomes hard to see the trees in the forest. Visualizing the magnitudes and effects of sorting can help understand what’s happening in the data sort of the head post.

  384. Andrew, sounds like a bit of nonsense to me.

    The number of papers with substantive errors (as opposed to spelling errors, equation transcription errors and the like) that I am aware of is a very small fraction of the total papers.

    Maybe its a lot more frequent in his area, I don’t know.

  385. Carrick,

    While I think he over does it a bit and he is talking in the medical field I think there are some lessons for climate science and especially data mining. To fit the first of his ideas to climate and trees (and this is my interpretation) : if there are a million possible sets of tree cores that could be fit to a temp series (and really there are many more) and we use a p=0.05 then we will have 5% or 50,000 false positives (that is tree sets that look like they fit but really don’t) — also is there any measure of the power or type2 error? That is of the trees we reject what is the risk of rejecting trees that really do fit but don’t appear to (I could find an estimate of power in Mann08)? If we go around testing and rejecting our tree cores on their fit to the recent record and then report this without independent verification of the choice and without a measure of the power then it is painting pretty pictures with no statistical basis — sure it might tell us where to look but it don’t prove very much.

  386. Carrick
    “all the while admitting no experience with the system!?”
    Umm no, you interpreted that. I have a lot of experience with the process of scientific publication.

    “But the fact is most scientists do go to at least one international conference each year, these typically happen 2-3 times a year. Most universities have grant monies set aside specifically for that, but I suppose you wouldn’t know that would you?”
    This is way too generalised. Many people can’t afford it when they don;t get their grant. It’s not some magic pile of money that allows all scientists to go to whatever conference they like. And what of those in poor countries? My wife doesn’t get too go to all the conferences she’d like to go to (and Australia isn’t poor). I think you are just assuming everyone has it as easy as you.

    ” and by the time the research gets accepted and published, the contents of the paper are often well known within the community.”
    well, this is interesting. So we have gone from all the players being aware of research by going to the conferences to the research being “often well known” by getting a pre-press of a publication. So this actually demonstrates that publication is more than just archiving, it’s a valuable distribution tool. In fact it’s the best tool science has ever had for the dissemination of knowledge.

    Why don’t you use your pre-press access to look at that paper I linked to earlier.

    “I think it’s wrong to not release your data once you’ve published it. (Hardly necessary to put that simple observation into a peer-reviewed publication though.)”
    It’s not a hard question is it? I’ll try again
    Do you think Biffra’s conculsions are correct?

    the release of data sounds like a load of whiney again.

  387. Carrick, at the theoretical end of my field research is almost entirely circulated by conferences and ArXiv preprints. Some key results are never formally published at all. Publications are much more important at the experimental end however.

    Re the Ioannidis paper, when I worked in biochemistry in a past life I came to the rough conclusion that 80% of published papers were “wrong” in some sense, though by far the most common way to be wrong was to be so misconceived as to be basically irrelevant. There’s an awful lot of rubbish out there. I seem to recall that even in Physics (where the average quality seems higher to me) about half of all published papers are never cited except for self citations by their authors

  388. Lucia: (Comment#22005)

    [quote]Tom Curtis–
    Lucia is a female name, and I am a woman.[/quote]

    Sorry for the misunderstanding.

    [quote]It is traditional to test statistical methods against noise. The rest of what you say reveals a profound lack of understanding of significance testing. It is traditional to test what a statistical method against “noise”. Even if trees do contain some signal, if that signal cannot be distinguished from noise, then you can’t conclude anything. Basically, if the same result would be obtained from noise and trees, then when you get the result, you must say, “This result may be ’signal’ or it may just be noise. I don’t know which.”

    That’s call lack of statistical significance.[/quote]

    I do not consider this to be an adequate responce to my criticism. If the sole criteria of the test is that it be tested against “noise”, then it is puzzling as to why you do not simply use the noisiest “noise”, ie, white noise. However, as you have already acknowledged in other comments, if you used white noise, or even red noise with a relatively low autocorrelation, your recipe for accidental cherry picking would not work. That means either you have cherry picked yourself by using red noise with a high autocorrelation, or you are forced to make the quite reasonable claim that different types of noise are relevant for testing different statistical techniques.

    If you make the later claim, you then need to defend yourself against my criticism which is that your choice of noise was inappropriate to test the technique. Your choice of noise means there are a relatively high proportion of data series having a characteristic shape which can be selected for by screening only a portion of the data series. However, this is not true of noisy data series in general. It would not be true of a noisy signal consisting of the superpostion of white noise on a sine wave, for example.

    If you constructed a large data set consisting of a variety of wave forms, including superposed wave forms, all with white noise superimposed on them, with the “signal” to “noise” ratio restricted such that there is an at least moderate autocorrelation (> 0.5), then used your cherry picking recipe and consistently got hockey stick shapes, then your criticism would be valid. But you and I both know that would be extremely unlikely. You can even include a portion of data series which are pure white noise, and pure red noise, and again you would be unlikely to consistently get hockey stick shapes. It is tests against this form of noise, however, which are relevant, for it is already known that the high autocorrelation in treering data series is a consequence of the climate sensitivity of tree growth.

    Because we know tree growth in sensitive to climate factors, we know there is a climate signal in the tree ring data series. What we want is a technique that will pick out that signal from the noise. Therefore, the valid statistical test is whether the technique used consistently picks out the signal from the noise, not whether it picks out the one shape available from data constrained to have only one significant shape, which is what the use of red noise does.

    Of course, even if a technique passes the test I recommend, it is still possible to argue that the signal picked out is not primarilly temperature, or global temperature; but that argument is not relevant to the validity of your cherry picking recipe as a test of statistical significance.

  389. steven mosher (Comment#22093)
    October 19th, 2009 at 9:36 pm

    “err duh, isnt the whole point of a tree/forest to redden the enviromental signals. or am I being stupid again?”

    This is the essence of all life, is it not? Stock-piling resources in preparation for a time of scarcity? Vegetative storage proteins are a tree’s stockpile.

  390. Nathan (Comment#22101)
    October 19th, 2009 at 10:47 pm

    “So why don’t you publish on why Biffra is wrong?”

    Briffa is wrong to have failed to disclose the sample size on his Yamal chronology. Briffa is wrong to have not disclosed his selection of Yamal over Polar Urals. Some sweet publicaiton that would make.

  391. Bender

    “Briffa is wrong to have failed to disclose the sample size on his Yamal chronology. Briffa is wrong to have not disclosed his selection of Yamal over Polar Urals. Some sweet publicaiton that would make.”

    Perhaps if you actually quantified the error, or showed what it meant?
    I think the fact that you acknowledge there is no publication in it shows that the ‘error’ isn’t actually important.

  392. bender
    I know you advocate science by blog, have you checked out the DelayedOscillator website? Should you be blogging over there? Don’t you need to adress all the claims on his(or her?) blog?

  393. Carrick (Comment#21982)

    My criticism is directed at Lucia’s recipe for cherry picking as it is applied to tree rings (and other proxies). It is not directed at criticisms of “hockey stick” graphs in general. In essence, the “hockey crew” claim to have a technique which will pick out the climate signal from noisy data. Lucia claims that their technique will pick out a characteristic shape from a very tightly constrained type of noise containing no data. IMO that is irrelevant. We know the data being tested is not pure noise. That means the correct test is, (a) will the the “hockey stick crew’s” technique correctly pick out the signal when there is a dominant signal; and conversely, (b) will it fail to pick out a signal when the data consists of a variety of noisy data, each with distinct signals.

    Lucia’s recipe for cherry picking fails to test either of these. However, if she ran it with data sets containing a dominant noisy signal to test (a); and a variety of noisy signals (including a some consisting of pure noise) with no overall statistical tendency to test (b); she would have a valid test. It is, however, by no means clear that the “hockey stick crew’s” techniques would fail such a test.

    There are, I agree, grave concerns about treating particular local proxies as “bell weather” proxies for global climate. There are also very reasonable concerns about treating a very small number of local proxies as a basis for determining hemispheric or global averages. I am sure even the “hockey crew” would agree that their techniques for studying historical climates are not perfect, just better than currently available alternatives. I am also sure that there are valid bases for disputing that last rider (without endorsing that conclusion). However, none of these were the basis of Lucia’s criticism in cherry picking article, and it was just that article that I was commenting on.

  394. Nathan (Comment#22116) October 20th, 2009 at 2:03 am

    actually those lacunae are enough for me to SUSPEND JUDGEMENT about the truth of briffas claims. His publication in my mind doesnt reach the level of a scientific claim. His testimony doesnt get in my court. sorry. If he were my student he would get an incomplete.
    lacuna in the published data, no source code = I suspend judgement

  395. What I find bizarre is that people keep complaining here that scientists keep utilising the same methods (or similar methods) to produce climate reconstructions; reconstructions that a lot here find dubious, problematic, cherrypicked etc. People keep claiming that the scientists should know better, that their training is bad or that they don’t understand statistics, and that they should learn how to do it right. BUT no one actually tries to get message to the science community using the most effective tool that’s available. Why don’t you do it? It’s not about appealing to authority, it’s about simply doing the work and presenting it in a readable format that is accessed by the target audience. To expect scientists to lounge around on blogs and read through the waffle (most of the interesting stuff is in the comments), to even find the ‘right’ blogs to read, is way too much. Publishing the work (hopefully) means that it’s not waffling and self-indulgent, but succinct and to the point – and if it’s wrong in the end, at least it is short.

    Why do people find it surprising that new reconstructions are done, and keep getting done despite the apparent errors in the logic, when you don’t formally outline the problems you see with them.

  396. Steven

    Well that’s fine if you want to play it like that, that would seem a good default position on any new science. But it would seem your friends here go beyond suspending their judgement. It seems they judge it very harshly.

  397. “What I find bizarre is that people keep complaining here that scientists keep utilising the same methods (or similar methods) to produce climate reconstructions; reconstructions that a lot here find dubious, problematic, cherrypicked etc.”

    Actually the complaint is that they don’t use certain methods. They don’t archive data; they don’t publish adequate descriptions of their novel methods; they don’t publish code; they don’t publish clear chronology retention criteria. They use novel procedures without documenting them first in peer reviewed statistical journals. basically, they are not doing science.

    ” People keep claiming that the scientists should know better, that their training is bad or that they don’t understand statistics, and that they should learn how to do it right. ”

    Actually they point out that Dr. Mann himself said that he was not a statistician. Further, I suppose if there were a journal dedicated to the misuse of stats one could get a paper out of detailing some of the issues here.

    “BUT no one actually tries to get message to the science community using the most effective tool that’s available. ”

    I’m not sure that you have actually shown what the most effective tool for this is. A simple study of the piltdown mann episode will show you that long past the time when others questioned the data, journals were STILL allowing in articles that ignored the hoax http://www.talkorigins.org/faqs/piltdown.html. So, I’d like to see some evidence in the form of data about ” the most effective tool.”

    “Why don’t you do it? It’s not about appealing to authority, it’s about simply doing the work and presenting it in a readable format that is accessed by the target audience. To expect scientists to lounge around on blogs and read through the waffle (most of the interesting stuff is in the comments), to even find the ‘right’ blogs to read, is way too much. ”

    Actually, the facts are against you on this one. Gavin and others read CA and other blogs regularly. In fact Gavin read CA and found out that steve had found a problem with Steig et al. Gavin re looked at the data, found the error that steve had written about. Then Gavin wrote to the BAS and alerted them to the error. Then Gavin claimed that “someone” had found the error independently of steve, Then BAS let everyone know this “someone” was Gavin himself. Sorry to raise facts about your hypothesis. Next.

    “Publishing the work (hopefully) means that it’s not waffling and self-indulgent, but succinct and to the point – and if it’s wrong in the end, at least it is short.
    Why do people find it surprising that new reconstructions are done, and keep getting done despite the apparent errors in the logic, when you don’t formally outline the problems you see with them.”

    It’s hard to get any more clear than “free the code.” You might not be aware, but through a concerted effort on the blogs we actually got GISS to take a step in the right direction by publishing their code. Writing papers are not the ONLY avenue to change or the best avenue, especially for the kinds of changes that I want or some of the changes that Steve wants. Where exactly does one publish a paper that demands that climate scientists publish their data and code. where? Well, we’ve used FOIA ( more effective than a journal article) used blogs ( more effective) and used complaints to journal editors ( more effective)

  398. Steven

    “It’s hard to get any more clear than “free the code.”
    This is a different issue. And as you pointed out earlier the apparent lack of ‘free-ing the code’ only makes you suspend your judgement.

    I think you have grossly overestimated the value of blogs.

    “Actually the complaint is that they don’t use certain methods. They don’t archive data; they don’t publish adequate descriptions of their novel methods; they don’t publish code; they don’t publish clear chronology retention criteria. They use novel procedures without documenting them first in peer reviewed statistical journals. basically, they are not doing science.”

    OK, so now we are going beyond suspending judgement into “They are not doing science”.

    If you have a problem with what they are doing you need to quantify it. All this moaning about inadequate descriptions is lame when you can’t quantify it. I’d say that it is actually your side of the debate that isn’t doing science, you’re attempting to make political arguments. You’re actually targetting the politics of science, and blogs are probably the best way to do that. But that’s not science.

    The point I am making is despite all the claims of broken hockeysticks etc they still do the proxy reconstructions… They keep churning them out. They’re doing science, getting published etc. All you guys are doing is whining on blogs.

  399. I think a key test of the value of ‘blog-science’ is how often you go to other blogs sites that disagree with your version of the debate and adress their discussion. Most of the discussion on CA revolves around published works, not blog posts on rival sites. Obviously published work has more impact else CA would focus on blog-posts.
    For example the Delayed Oscillator website had lots of interesting things to say about the Yamal whoopsie. Shouldn’t CA be looking at those claims and auditing them?

  400. Between Lucia, Jeff id, and most certainly the great Canadian Steve McIntyre, the hockey team is really feeling the heat. As a Canadian, and former fair hockey player, I must say, this isn’t a surprise, most international hockey teams end up bleeding , dispirited and defeated when playing against Canada.

  401. “mainstream climate scientists boycott these highly visible online discussions or, when they occasionally break the boycott, do little more than moon the discussions with adolescent drive-by comments ”

    Oh, come on. You can’t traffic in snark and then complain when it’s returned in kind. Even if you feel that climate guys “started it first” it comes off as unprofessional and erodes CA’s credibility, which, being a blog and new and all, is not substantial to begin with.

  402. Tom Curtis

    I do not consider this to be an adequate responce to my criticism. If the sole criteria of the test is that it be tested against “noise”, then it is puzzling as to why you do not simply use the noisiest “noise”, ie, white noise.

    You use noise with the appropriate features. In fact, if we always used white noise, we would now consider HadCrut to show statistically significant cooling since … well certainly this decade.

    The correct question to ask about the red noise is– if we the trees are “noise”, is it reasonable to consider the noise “red” (or something else). I don’t know. It’s almost certainly not white. If the temporal autocorrelation for the trees is “white” it can’t be a good proxy! After all, temperature exhibits strong temporal autocorrelation which is not white.

    So, the noise must be something other than white. Is red the best choice for trees? I don’t know. This is a “toy” example. But white is a poor choice. Is the monthly temporal autocorrelatin “too high”…. well, I don’t know. But, if you don’t detrend (which is necessary for this test) the temporal autocorrelation for temperature anomalies is high. So good temperature proxies would have pretty high temporal autocorrelation.

    (On the toy example issue, given that treerings don’t form once a month, I suspect that you can’t really even use monthly data– you hve to use annual average data.)

    Because we know tree growth in sensitive to climate factors, we know there is a climate signal in the tree ring data series. What we want is a technique that will pick out that signal from the noise. Therefore, the valid statistical test is whether the technique used consistently picks out the signal from the noise, not whether it picks out the one shape available from data constrained to have only one significant shape, which is what the use of red noise does.

    Yes. And my point is only that the method needs to be tested against appropriate synthetic noise, and the tests need to be fair.

    I wrote thispost because The Lorax at CA (and some on threads elsewhere) don’t even recognize that a) screening against temperature can even potentially cause problems and b) because it does you need to be careful and do any test appropriately.

    I did a “toy” case because it illustrates the potential problem. I’ve never said this applies to any particular paper. I don’t read the dendro papers much. To many picky details to track down to make it worth my time.

    (BTW: I”m offline alot after taking Dad to a retirement home over the past two weeks, my father-in-law is having chemo tomorrow. So, “family stuff” is sucking the time.).

  403. “It’s not about appealing to authority, it’s about simply doing the work and presenting it in a readable format that is accessed by the target audience.”

    It is about appealing to authority. You Warmers claim that critcisms can be rejected based on the fact that they haven’t been accepted by members of the Warmer’s Club. They don’t have to be.

    Imagine for a moment (Analogy Time) the Kansas City Royals and the New York Yankees play a baseball game in London (a neutral site). The Royals win.

    Yankee Fan complains: “Yeah, but the Royals haven’t proven anything. They are nothing until they come to Yankee Stadium and beat the Yankees on their home field. (where the fences are closer, the baseballs are slicker, our players are more comfortable, etc.)”

    This is comparable to what is happening now- The Warmers are getting beat at a neutral site (the internet) so they complain that the game isn’t rigged in their favor. Bring it to Our House to prove yourself. Appeal to Our House. That’s all this is.

    Andrew

  404. lucia (Comment#22127)
    October 20th, 2009 at 6:15 am
    Trees are far redder than even the reddest environmental inputs. Could I list papers? Yes, about 40 of them.

  405. Nathan, the ugly truth is you won’t be one of the leading researchers in your field if you can’t go to international conferences at least occasionally (once ever other year at worst). There is no way for peer-reviewed publications to substitute for that experience. You have a lot of opinions about a matter that you don’t personally delve into, and there is no way that even living vicariously through you wife is going to substitute for that.

    Live with your ignorance, it’s no issue to me.

  406. Jonathan:

    I seem to recall that even in Physics (where the average quality seems higher to me) about half of all published papers are never cited except for self citations by their authors

    That’s true. However, being uninteresting to the community is not the same thing as being wrong, which is what I was responding to. And I’ll grant that medical research tends to be flawed much more often… and there’s a reason for that. 😛

    However, if anything this speaks against Nathan’s silly argument that the only or even chief way that scientists communicate information is via peer reviewed publications. I know this is false because I know plenty of scientists who don’t routinely read the journals (this is much more likely to happen if they are at the top of their field of course).

  407. bender:

    Trees are far redder than even the reddest environmental inputs. Could I list papers? Yes, about 40 of them.

    I’d like to see one or two if you don’t mind.

  408. OK this is all I have time for, Nathan:

    This is way too generalised. Many people can’t afford it when they don;t get their grant. It’s not some magic pile of money that allows all scientists to go to whatever conference they like. And what of those in poor countries?

    How you are going to do research in my area (or any areas I’ve worked in the past) without funding is beyond me. Fortunately, Universities do provide grant monies to promising researchers with no research funds… short term.

    If you don’t keep funded long term, you won’t get tenured. That’s as important as peer reviewed publications. And if tenured but no funding, you will gradually end up concentrating on teaching. It’s the way of the wild in the experimental research community.

    My wife doesn’t get too go to all the conferences she’d like to go to (and Australia isn’t poor).

    Nice waffle. Want some syrup with it? Seriously I never said we go to “all of the conferences”, just “enough” of them. Even if you have all the funding in the world, if you are a serious researcher, you can’t afford taking that much time away from your work.

    I think you are just assuming everyone has it as easy as you.” and by the time the research gets accepted and published, the contents of the paper are often well known within the community.”

    I know that to be a fact. You don’t because you have no personal experience with research.

    well, this is interesting. So we have gone from all the players being aware of research by going to the conferences to the research being “often well known” by getting a pre-press of a publication. So this actually demonstrates that publication is more than just archiving, it’s a valuable distribution tool. In fact it’s the best tool science has ever had for the dissemination of knowledge.

    Conferences (and then preprints) are the most effective methods for disseminating the ideas. Papers (peer reviewed or otherwise) are important for disseminating the knowledge in greater depth. Peer reviewed pubs act primarily as a “archive” for the knowledge, not the means by which it is first made available. The INTERNET is the chief distribution tool these days.

    These days for experimental research, however, the web is steadily becoming a more and more important tool. Archiving of much larger data sets, access to full code a

    Why don’t you use your pre-press access to look at that paper I linked to earlier.

    Pfft.

    It’s not published in a peer reviewed journal yet! What me peek at anything that’s not peer reviewed!?!!! 😛

    Do you think Biffra’s conculsions are correct?

    Not my monkey, and a diversion from this other “bizarre” topic. I can complain about his lack of sharing of published data, because that is a universal characteristic expected of all researchers.

    All I have time for.

  409. Carrick,
    Ok. But only if you first read chapter 1 of Hal Fritts “Tree rings and climate”.

  410. Tom Curtis,

    In the examples I provided, part1 & 2 address the concerns about how ‘red’ the noise should be. In part 1 – actual data. In part 2 shweingruber fit AR1 data was used.

    It doesn’t matter because in the other two links given above any amount of noise simply results in ever more signal distortion. Therefore the only way you can construct a correct historical signal using correlation sorting is by the assumption that the trees that pass are actually perfect noiseless thermometer trees otherwise the noise is guaranteed to bias the signal. There is simply no way around it.

    This link shows what happens as gradually more noise is added into the recostruction. You might need to read part 2 linked above to figure it out.
    http://noconsensus.wordpress.com/2009/08/26/more-on-cps-signal-distortion/

    After that, I have difficulty understanding how people could even consider that correlation sorting ‘might’ work. Actually, it never works if there is any noise on the signal.

  411. Nathan (Comment#22124)
    October 20th, 2009 at 4:25 am
    “Most of the discussion on CA revolves around published works, not blog posts on rival sites. Obviously published work has more impact else CA would focus on blog-posts. For example the Delayed Oscillator website had lots of interesting things to say about the Yamal whoopsie. Shouldn’t CA be looking at those claims and auditing them?”
    .
    McIntyre was an invited expert reviewer of IPCC AR4. That is why he tries to get the blog to focus on literature used by IPCC. Anyone who thinks there is great stuff out there in the blogosphere is welcome to bring it to the attention of the CA audience – iff its relevance to IPCC-cited literature can be made crystal clear.

  412. bender and layman lurker, thank you. This question has bugged me for a long time.

    “Please explain why you think the argument about selection bias is settled.”

    Because of Lucia’s demonstration, Lubus’s demonstration, countless others.

  413. Jeff Id,
    Great posts and blog too, I think. I am going to have to work on my stats chops to read it though.

    Nathan,
    You might want to check out WUWT http://wattsupwiththat.com/2009/10/19/proof-that-media-is-hyping-agw-shamelessly-with-asking-basic-questions-like-did-you-check-the-lake-for-ddt/ to see your Baffin Island paper ‘whacked’. It seems that the disapearance of those midges, the smoking gun in climate change, they point to, happened at about the same time DDT came into widespread use there. I didn’t look at the other one, sorry, the first was such a laugher.

    One thing the interweb has time for, its “whack-a-mole” and they charge people to play the real game because… it’s fun.

  414. TJA, I think that it is important to keep in mind that it is almost impossible to determine the extent of selection bias in any given chronology or its effects (much of the reason is the lack of archiving and metadata). However, IMO It is reasonable to assume there is at least some “noise” captured during calibration. It also seems clear that in many cases the methods fail to lay the proper foundation for valid conclusions wrt this type of bias (among others). Proper psuedo data cases help to establish base thresholds which authors should test against. Check out Jeff’s link above which shows the impact of progressive s/n ratios.
    –
    BTW IMO there are many other issues related to data quality, data processing, and analytical methods which can have much the same consequence – noise in the reconstruction period.

    Carrick, here is link to a paper by Dr. Koutsoyiannis (check out his site for other links). I like this paper because it actually uses a Graybill standardized tree ring series from Mammoth Creek, Utah as a case example of “long term persistance” which goes well beyond AR1.

    http://www.itia.ntua.gr/getfile/511/1/documents/2002HSJHurst.pdf

  417. Nathan (Comment#22123) October 20th, 2009 at 4:19 am

    ““It’s hard to get any more clear than “free the code.”
    This is a different issue. And as you pointed out earlier the apparent lack of ‘free-ing the code’ only makes you suspend your judgement.”

    I would argue more than this. I would argue that it is not rational for you to exercise your judgement without access to the code. In short, there can be no rational basis for consensus. So, yes it makes me suspend my judgement which means there is room for doubt, rational doubt. And it the face of that doubt I see no need to change policy or change my behavior and anyone who is arguing that we must change has no rational footing. Further, it is not a different issue. The issue is and always has been the scientific method whether you are talking about foundational methodological principles or methods further up the chain of evidence. If the authors make mistakes at the foundational level then that certainly calls into question their ability, judgement and understanding. So it’s not a different issue. It’s the same issue at a different level.

    “I think you have grossly overestimated the value of blogs.”

    Well, I asked you for evidence that journals were the most effective place to make the changes that we argue for. You supplied none.
    I give you an F either on reading or on debate, take your choice.
    In turn I gave you evidence that blogs, FOIA, and complaints to editors do work. On balance I would say that if you provide no evidence and I provide evidence that rationally speaking you should concede the point. Instead, you counter with a waffle.
    You think I over estimate the effectiveness.You are wrong. I think you underestimate it.

    ““Actually the complaint is that they don’t use certain methods. They don’t archive data; they don’t publish adequate descriptions of their novel methods; they don’t publish code; they don’t publish clear chronology retention criteria. They use novel procedures without documenting them first in peer reviewed statistical journals. basically, they are not doing science.”
    OK, so now we are going beyond suspending judgement into “They are not doing science”.”

    No, that is the basis of suspending judgement. Do you NOT understand the structure of argumentation?

    “If you have a problem with what they are doing you need to quantify it. All this moaning about inadequate descriptions is lame when you can’t quantify it.”

    Well I can and have quantified it on several occasions. In particular I quantified it for GISS with regard to their analysis of global temperature. {There are even guidelines for doing reproduceable results that have been published which I passed onto them } . I will give you one simple example. In hansens last paper on the global temperature index his text indicated that 5 northern california sites were removed from the dataset because of anomalous cooling in the early 20th century. That is all the paper said.
    I asked for a description of the statistical test that was performed to justify this removal. No answer. Finally, after the code was released one could see the documentation. What did I find? The description said the sites were removed “by hand” after a subjective review. “they looked too cool” Another example, would be the retention criteria for tree ring series. That is, what kinds of screening tests are done ( core counts, EPS, Correlation, etc) prior to analysis. There is no published canonical method or proceedure, so various papers use various standards. take EPS for example. a figure of .85 has been used ( but not verified or validated ) but as you read through literature you can find people using, .8, .7 etc.

    ” I’d say that it is actually your side of the debate that isn’t doing science, you’re attempting to make political arguments. ”

    Let’s see. I think the best counter example to this is the starbucks hypothesis which McIntyre supplied evidence for. ( that’s a joke, but you go look it up) It is trivially true that pointing out the methodological flaws in a science process is and can never be construed as science. If a scientist says that 2+2=5, and I point out that 2+2=4, then it is trivially true that I am not “doing science.” I am doing math. Further your argument does not follow.
    Just because I am not doing science, does not imply that I am “doing politics”. Finally, when A climate scientist argues, as hansen has, that we need to change our behavior or face a catastrophe is he “doing science?” Climate science has on it’s own politicized the discussion. Finally, you dont EVEN KNOW what my side of the debate is. I am a luke warmer. That means I believe in AGW theory. I believe that radiative physics is sound science. I don’t believe that paeloclimatology is there yet. Precisely because they engage in certain behaviors. Climate modelling does somewhat better. Politically, I believe that dumping GHGs into the enviroment is not a well thought out long term strategy, however, I refuse to be bullied or scared into drastic action.

    “You’re actually targetting the politics of science, and blogs are probably the best way to do that. But that’s not science.”

    Trivially true or pointless.

    “The point I am making is despite all the claims of broken hockeysticks etc they still do the proxy reconstructions… They keep churning them out. They’re doing science, getting published etc. All you guys are doing is whining on blogs.”

    I will make this simple. Until they release that data and the code they are not “doing science” They are publishing words and pictures. Other people look at those words and pictures and call it “science” but science is not a publication. It’s a process, a set of behaviors. key to that process is the ability of others to “do the same thing” Look. I tell you this. If you drop a mass in a vacuum it will fall at a certain rate. I describe what I did, show you my data, and explain how you can do the same thing. Why? well partly to remove things like observer bias. You don’t believe me? fine, here are the instructions to repeat my behavior ( my science-ing) and remove “me” ( and my politics, my hair color, etce ) from the equation. If I don’t do that, then I am not doing science. If you criticize me, you are not doing science either.
    And, there is more than whining going on in the blogs. At CA for example Dr. Craig Loehel submitted his analysis of proxies to the community for their comments. They criticized his paper. Another professor with expereince in time series statistics offered his help. Together they co authored a paper and had it published. After publication, other blogs found errors. Those were acknowledged and corrected. Your view of blogs is overgeneralized. you need to be more specific and scientific in your criticism.

  418. Nathan,

    “I think a key test of the value of ‘blog-science’ is how often you go to other blogs sites that disagree with your version of the debate and adress their discussion. “

    Exactly so. And how often do you see RC address issues on sites that disagree with them? How many times do you read litanies of posts to RC ,questioning the science that they are promoting, that never get published?

    Get real Nathan. And follow Mosher’s advice

  419. Nathan argues for the primacy for journals, elite science, peer review and sneers at blogs which raises a question as to why he is here at all.

    Blogs such as this one have become important in climate science for the simple reason that the normal “establishment” is virtually corrupt and cannot be counted upon to vigorously examine methods, assumptions and findings. That a dog’s breakfast like Mann’s hockey sticks have the iconic status they do is a pretty good indication that ideology has trumped empirical rigor and the extent to which it is alarmism über alles.

    I find that a good secondary proxy of the strength of a “denialist” argument is how it is treated on the alarmist blogs. Tamino rips weak mathematical presentations with enthusiasm but if he simply ignores or issues a dismissive harrumph to something that is widely cited, I infer that he could not find a handle.

    Similarly, I take careful note of RealClimate’s varying levels of snark, the number of complaints on non-alarmist blogs about comments deleted or users banned on particular threads, whether the RealClimate author offers relevant detailed reasoning or just offers eye-rolling ad vericundium references and –more snark.

    The climate scientific “establishment” has largely rolled over for the additional funding, media attention and policy influence that alarmism generates and which political correctness enforces.

    That so much insightful and detailed critiques come from outside that establishment should not be surprising but should be celebrated. Climate Audit , the Blackboard et al are to climate science what the Irish monks were to the Dark Ages — stubborn outposts of methodological civilization during a time of sweeping ideological conquest. (Was that over the top? I can never tell…)

  420. Carrick

    “Nathan, the ugly truth is you won’t be one of the leading researchers in your field if you can’t go to international conferences at least occasionally (once ever other year at worst). There is no way for peer-reviewed publications to substitute for that experience. You have a lot of opinions about a matter that you don’t personally delve into, and there is no way that even living vicariously through you wife is going to substitute for that.

    Live with your ignorance, it’s no issue to me.”

    Oh so now you only have to go occassionally? Look you made the stupid claim that everyone attended every conference. Then changed it to claim that pre-press helpd spread the word. The fact is, publishing your work is important. Why don’t you just admit your initial claim was wrong?

    “How you are going to do research in my area (or any areas I’ve worked in the past) without funding is beyond me. Fortunately, Universities do provide grant monies to promising researchers with no research funds… short term.”

    This is because you think only in your narrow experience. My wife publishes regularly and does not work for a university, she is an expert in Western Flower Thrips, and works for the State Government here. Funding cuts are very common. This does not mean she is no longer doing research, but certainly means she can’t go to every conference. Your glib claims are nonsense and I think you do understand the importance of publishing.

  421. Dave Andrews

    “Exactly so. And how often do you see RC address issues on sites that disagree with them? How many times do you read litanies of posts to RC ,questioning the science that they are promoting, that never get published?

    Get real Nathan. And follow Mosher’s advice”

    Dave how often does a skeptic actually publish? There’s not much for them to look at, and as soon as someone publishes something they look at it. They do address blog poss, but that’s because they are fighting a political fight too.

    If publishing is not important why does CA only focus on publishd work? Tell me does CA go around the intertubes looking for all the posts on blogs that disagree? I expect to see some posts about Delyed Oscillator then… Come on, blog science is important, you need to address all those blogs… Hey just start at ScienceBlogs.

  422. George

    “Nathan argues for the primacy for journals, elite science, peer review and sneers at blogs which raises a question as to why he is here at all.”

    You just made this up.

    Here is the real quote:

    “Nathan argues for the primacy for journals.”

    You can do whatever you like on your blog, but if you want to change the way proxy reconstructions are done you’re going to have to publish on it. The LEAST you can do is publish. Why is it so hard? Why do you labour over this? Why do you dodge the inevitable criticism? Why do you have to pretend that publishing is not important to justify the lack of publication?

  423. George

    “The climate scientific “establishment” has largely rolled over for the additional funding, media attention and policy influence that alarmism generates and which political correctness enforces.”

    Yes, becuase they are SOOOOO wealthy.

  424. Steven Mosher

    If blogs science is so good, why does CA focus only on Published work? Why don;t they look at other ‘good blog science’ sites.

    As to your 2+2 = 4 claim. This is simplistic nonsense, the mathematics involved in proxy reconstructions is not trivial. If you can demonstrate that proxy reconstructions are flawed why do you hide it at your blog? Why wouldn’t you publish it? I think I know why, and that’s because there’s nothing really to publish. The ‘problems’ are trivial. You gave your JHanson example, why didn’t you add those data back in and quantify the difference? Why did you just stop at that point and laugh? Why don;t you follow it through to it’s logical end point.

    The longer you postpone publishing your results, the longer you postpone quantifying your results, the longer you avoid following your work, the less scientists will take notice. Heck there isn;t even a summary page for all th published work CA has demolished. If Steve won;t publish he could at least make up a summary outlining exactly the problem he has with various works. At it stands the site is a dogs breakfast. Look at the current post there on Yamal substitution. Are you saying he couldn’t publish a paper using that? If he outlined his argument using Polar Urals and published it then the dendochronologists would actually have something to debate.

    “I will make this simple. Until they release that data and the code they are not “doing science” They are publishing words and pictures.”
    This is so disingenous, you know that there are a lot of issues at play here. This is a cheap excuse for argument. What happened after GISS released it’s code? Write any papers on what is wrong with it? Anything happen? Or did you just discuss on your blog how awful it is without doing anything? Did you find the hidden ‘make it warmer smoking gun’?

    What does being a Lukewarmer mean anyway? I’ve not seen any published work about luke-warming, I;ve not seen any analysis. Surely if you belive in luke-warming you have some sort of sensitivty estimate or forcing value for CO2 or ANYTHING AT ALL that you can publish about your theory. Where do you get your luke-warmer beliefs from?

  425. Nathan, I did say one international conference every two years or so. You’re acting like an ass.

    Have the floor.

  426. Bender

    “McIntyre was an invited expert reviewer of IPCC AR4. That is why he tries to get the blog to focus on literature used by IPCC. Anyone who thinks there is great stuff out there in the blogosphere is welcome to bring it to the attention of the CA audience – iff its relevance to IPCC-cited literature can be made crystal clear.”

    How generous of you. Ad here I was thinking that you need to go out and find new science (or as Carrick says attend conferences). But no, actually it needs to be brought to you.
    I already told you about DelayedOscillator. There, I gave you something to look at. Or does it need to be a link at CA? What are the rule, pray tell.

    “That is why he tries to get the blog to focus on literature used by IPCC.”
    Ok , so lets find out what proportion of literatue used in AR4 has been audited? Are there many papers he agrees with? Or is there some focus on proxy reconstructions? Where are the summaries on what is wrong with various papers? How will a wandering dendrochronologist, who wishes to learn about the CA way, source the gems of wisdom there?

    “iff its relevance to IPCC-cited literature can be made crystal clear.”
    Ummmm Was Yamal made crystal clear in this respect? The Delayed Oscillator blog addresses Yamal specifically. So… I guess it must be.

  427. Nathan, your Carbon Footprint® on this thread alone is astounding! How much closer have you brought us to the Tipping Point®? Please constrain yourself – do it for the midges, man!

  428. Carrick

    “Nathan, I did say one international conference every two years or so. You’re acting like an ass.

    Have the floor.”

    Your words:”In my area(s), yes, we do all attend the same conferences.”

    Get over it. This was not an important issue yet you continued to say I was ignorant or arrogant. I even asked you to clarify if you expected everyone to attend and you said yes. I’d have to say the ass here is you.

  429. TerryMN

    It seems the ‘analysis’ on WUWT seems to stop right at the critical moment. In the comments they say it has nothing to do with DDT. What a stunning rebuttle of something they haven’t even read. I assume, Carrick, that the went to the conference and didn’t need the paper.
    What a stupid post.

  431. Nathan, I did write:

    But the fact is most scientists do go to at least one international conference each year, these typically happen 2-3 times a year. Most universities have grant monies set aside specifically for that, but I suppose you wouldn’t know that would you?

    And to clarify, by international conference, I mean conferences that have significant international attendance, not necessarily conferences you attend on foreign soil.

    If you want, we can vote and see who people think is acting the ass.

  433. Nathan – “midges” was topical – don’t get hung up on them. Please feel free to substitute polar bears, Maldives, glaciers, ice caps, penguins, the great barrier reef, ski resorts, Tuvalu, pikas, Venice, salamanders, walrus, caterpillars, coral, frogs, or hell – as The New Scientist put it 5 years ago: “Global warming may drive a quarter of land animals and plants to the edge of extinction by 2050, a major international study has warned.” – pick just about anything – but keep it on topic of your many, many (dangerous!) posts and excessive use of electricity.

    In summary, I cannot believe that, given your viewpoint on the coming doom and the current Climate Crisis® that you can be so cavalier and irresponsible, knowingly spewing poisonous C02 with your myriad of posts, and threatening the lives of millions of species. How do you sleep at night?

  434. RyanO–
    I don’t read anyone pestering me to knit them slippers or telling us about their burning desires to impregnate Sarah Palin. So, I don’t think the ‘TCO’ level has been breeched.

  436. I had a couple a brief exchanges with TCO here a while ago. I remember them pretty well, though. There were two points that he, for some reason, decided it was OK for him to state for the record. (Otherwise he just slip-slided and pretzeled around in his responses)

    1. He didn’t believe in AGW.

    2. He didn’t beat his wife.

    I didn’t know what to take from the incident, and I’m still in a state of wonder about it. Maybe Lucia has it/them in the archives. 😉

    Andrew

  437. Nathan starts pwning (Isn’t that what you kids say?) this thread and Carrick begins his usual insults and Ryan O wants filters.

  438. Nah. You get pwned when your adversary hits you with a death blow (metaphorically speaking in this case 🙂 ) out of nowhere. The word for repetitive posting is not “pwning”, dear Boris. It’s called “spam”.
    .
    😉
    .
    BTW . . . I was referring to the “repeat post” filter, not the one that makes someone go away entirely. 🙂

  439. Boris,

    Don’t you have Peer-Reviewed Journals to read? Or have you read them all already? Why are you wasting your precious time reading this Bloghorea? 😉

    Andrew

  440. Boris, I think people missed the point I was making.
    You can’t substitute blogs for published work.
    Blogs are fine for discussions and trying out views etc. But in the end you can’t substitute a blog post for publishing. Although, I can see an advantage in at least making succinct summaries of problems with papers.

    RyanO – hey you shouldn’t be worried, at least you’re attempting to quantify an error and publish it. Good work I say.

    TerryMN, I think you are just mocking something you don’t understand. What exactly is wrong with the paper? Have you checked whether DDT was sprayed near the lake, do you know the effect of DDT on the chrimo-whatchits (I don’t remember the insect name)?

    Carrick – the problem is that you are an ass for changing your argument and calling me ignorant and arrogant, doesn’t matter what others here think.

  441. RyanO

    “The word for repetitive posting is not “pwning”, dear Boris. It’s called “spam”.”
    Actually we were engaged in a discussion.

  442. Nathan, you’re absolutely correct. I’m mocking you, and I don’t pretend to understand you via your ramblings. Kindly start trying to make some sense (bonus – download a spell checker). Thanks, and have a great day!

  443. “You can’t substitute blogs for published work.”

    Yes you can. If the content of the blog is better than the content of the paper… go with the blog. If the paper is better… go with the paper.

    Is it really that difficult to grasp?

    Andrew

  444. Nathan, I didn’t change my argument. You simply failed to comprehend it.

    Boris, like everything you say isn’t an insult. LOL.

  445. Carrick, you seem to have a bad memory, I show you below how your argument moved. initially it was all we need to do is attend the conferences, then it became about pre-press and mostly attending conferences, and then was basically a bunch of lines about how ignorant I was blah blah blah.

    Carrick:
    Seriously though most science gets done away from the peer review process. Peer review is where you generally archive your work for antiquity, not where you break new ground—or at least not often, I’ve had a few occasions where this has happened with my published papers creating a bit of a stir, but generally people have heard your work ad nauseam at conferences before it ever gets published.

    Nathan: What a strange thing to say. Of course science gets ‘done’ away from Peer Review, you do the science first then submit it.
    As to ‘archiving your work for antiquity’ – well that’s a very strange thing to suggest. How do you expect people from other countries to know what you’re doing? Does everyone have to attend your conferences

    Carrick:
    In my area(s), yes, we do all attend the same conferences.
    What’s the BFD?

    Nathan:
    Seriously? everyone in your field attends the same conferences? That’s a rather unique situation. The BFD is that most science types can’t attend every conference – often because they live in different countries – so your ‘publishing for archival purposes’ line may apply to you specifically but hardly applies universally.

    Carrick:
    Over 25 years, I’ve done research in a wide range of areas, and it’s been my experience it works this way in practice in all of those. Nothing unique about it, any more than there is anything unique about your smug but ignorant arrogance.

    Nathan:
    I think you are the one generalising, and for whatever reason pretending that publishing is little more than ‘archiving’.

    Carrick:
    Surely as wise a one as you is published, yes?

    Nathan:
    Ha ha! Look I never claimed to be wise, buthey you can think whatever you like.

    Carrick:
    Nathan, I take that as a “no” as in you haven’t published.
    So seriously, you are here trying to lecture us on how science works, from a completely ignorant perspective?

    Nathan:
    No, I am not ignorant of the process. I am not lecturing you on how science works. You made a claim that publishing was basically for archiving. I think we both know that that is not a sensible claim.

    Carrick:
    Nathan, it is a sensible claim because it is true. And indeed you keep “lecturing” me on the process because you keep telling me what I know to be true, is false, all the while admitting no experience with the system!?

    But the fact is most scientists do go to at least one international conference each year, these typically happen 2-3 times a year. Most universities have grant monies set aside specifically for that, but I suppose you wouldn’t know that would you?
    Beyond that, you seem to be unaware of preprints, e.g., the arXig.org (what we used to call the XXX archive back when it was hosted at LANL). Again, these get widely circulated and discussed (as do, I might mention, the PowerPoint slides, which get disseminated to people who couldn’t make the conference), and by the time the research gets accepted and published, the contents of the paper are often well known within the community.

    Nathan:
    Umm no, you interpreted that. I have a lot of experience with the process of scientific publication.
    This is way too generalised. Many people can’t afford it when they don;t get their grant. It’s not some magic pile of money that allows all scientists to go to whatever conference they like. And what of those in poor countries? My wife doesn’t get too go to all the conferences she’d like to go to (and Australia isn’t poor). I think you are just assuming everyone has it as easy as you.

    …well, this is interesting. So we have gone from all the players being aware of research by going to the conferences to the research being “often well known” by getting a pre-press of a publication. So this actually demonstrates that publication is more than just archiving, it’s a valuable distribution tool. In fact it’s the best tool science has ever had for the dissemination of knowledge.

    Carrick:
    Nathan, the ugly truth is you won’t be one of the leading researchers in your field if you can’t go to international conferences at least occasionally (once ever other year at worst). There is no way for peer-reviewed publications to substitute for that experience. You have a lot of opinions about a matter that you don’t personally delve into, and there is no way that even living vicariously through you wife is going to substitute for that.
    Live with your ignorance, it’s no issue to me.

    And on it goes, despite me giving evidence that people conduct research without being able to attend every conference. You won’t admit you’re wrong so you keep going with this pointless argument and accusing me of being ignorant or arrogant or an Ass.

  446. Nathan, where did I say “every” conference? Perhaps you can point that out?

    Are you trying to be dense?

  447. Nathan/Carrick–
    You are now arguing about the arguments in the thread instead of substance. I think it’s time to stop. (I’m a bit too busy to moderate much. Last two weeks were my father, the next two are Jim’s father who goes into Chemo tomorrow. I may have to set “autoclose” to 4 days because after too long, bickering always ensues.)

  448. Carrick

    You are being evasive.

    Nathan: Does everyone have to attend your conferences

    Carrick:
    In my area(s), yes, we do all attend the same conferences.
    What’s the BFD?

    Nathan:
    Seriously? everyone in your field attends the same conferences? That’s a rather unique situation. The BFD is that most science types can’t attend every conference – often because they live in different countries – so your ‘publishing for archival purposes’ line may apply to you specifically but hardly applies universally.

    Carrick:
    Over 25 years, I’ve done research in a wide range of areas, and it’s been my experience it works this way in practice in all of those. Nothing unique about it, any more than there is anything unique about your smug but ignorant arrogance.

    Quite clearly you are implying that you needed to attend every conference, and then when I made the point that this was difficult rather than saying that you agreed with me you kept on arguing. Why did you keep on arguing if this wasn’t your argument? Why are you still arguing now?

    And in summary you still think I am smug, ignorant, arrogant, dense, and an ass? Bravo I say.

    “You have a lot of opinions about a matter that you don’t personally delve into, and there is no way that even living vicariously through you wife is going to substitute for that.
    Live with your ignorance, it’s no issue to me.”
    And this is a great comment, I mean you have no idea how involved I am yet you feel free to call me ignorant. How do you know I am not personally involved in the publication of science?

  449. Nathan, with all due respect, the host has asked for this discussion to leave this thread. You’ll have to find something else to be argumentative about. >.>

  450. “burning desires to impregnate Sarah Palin”

    eck…eck….EEECCCKK.

    Worse than the frozen rabbit fuel story. Good that you applied that filter.

  451. “Nathan, with all due respect, the host has asked for this discussion to leave this thread. You’ll have to find something else to be argumentative about.”

    Hear, hear.

  452. Lucia,

    I guess there is a lot on your plate these days. You and Jim are a tribute to your fathers, who must have been giants. Prayers and best wishes. Don’t know how, but if I can help, ask…

  453. Nathan:

    I’m starting from the bottom of the thread and working up, so my apologies in advance if I seem to be picking.

    “…feel free to call me ignorant. How do you know I am not personally involved in the publication of science?…”

    Nathan, we don’t know. You are hiding behind an anonym. So are a lot of others on both sides of the issue. If you have relevant experience, come out and tell us.

    One small clue to consider… anyone who extensively posts in the middle two weeks of October is probably not an active academic.

  454. Robert

    As I have said here many times I am an editor. I do technical editing of geological work. I am very aware of the problems of getting things published, and with how and why people publish.

    ‘anyone who extensively posts in the middle two weeks of October is probably not an active academic.”
    No, I am not… but I might be soon. Wish me luck.

  455. Robert

    “One small clue to consider… anyone who extensively posts in the middle two weeks of October is probably not an active academic.”

    I find this interesting, why would the date have anything to do with it? Perhaps in the USA this is significant, not sure what it has to do with anything else. My busiest time is from March to July.

  456. Nathan, I think Robert was pointing out that in most parts of the world the academic year begins in September and that academic teaching loads are at their greatest in the period from September to March, thus they have relatively little time to spend on posting.

    My more substantive point relates to your comments (and many others) about why these criticisms are not published in the peer reviewed literature. My view on this is that such papers are simply not accepted by journals who are only interested in publishing novel material. (in a similar vein it is almost immpossible to get papers published which demonstrate the absence of effects of substances, whereas the most trivial eefects are readily published)

    If Mr McIntyre was an active dendroclimatologist then he would certainly be publishing his own work and heaping criticism on his peers, but he isn’t. He is simply exposing serious statistical holes in published papers that should have been identified in the peer review process, but weren’t.

    Since he is also exposing some of the fundamental assumptions of the argument to be based on flawed methodology, no one want’s to listen. The credibility of too many people is now under threat.

    Consider that at Copenhagen world leaders are seriously considering the expenditure of trillions of dollars to control CO2 emissions based almost solely on the proposition that the current warming phase is “unprecedented” and is directly due to human activity. If the warming is not unprecedented and there are questions abouut the CO2 forcing built into the GCMs then it may be that the “Emperor has no clothes”, but many people are scared witless that this may be true and are fearful of the outcome to their reputations.

  457. Arthur

    Thankyou for your polite reply 🙂

    “He is simply exposing serious statistical holes in published papers that should have been identified in the peer review process, but weren’t.

    Since he is also exposing some of the fundamental assumptions of the argument to be based on flawed methodology, no one want’s to listen. The credibility of too many people is now under threat. ”

    Well, how do we know that? How do we know that Steve is right? I don’t agree he has demonstrated serious flaws at all. For example in his latest blog post on Yamal substitution, a wandering dendrochronologist (actually someone who had used that data) ventured into CA, told him why they had chosen the way he had. He also explained why he wouldn’t post there regularly but said he was in communication with Steve. Why would we believe Steve over Rob Wilson? Rob seemed to explain fairly why they had chosen Yamal over Polar Urals.
    If Steve thinks he has shown statistically that Dendrochronogy is seriously flawed, then he actually owes it to science to write that up and publish it. He could get it published in a statistics journal. I mean imagine the impact that would have, you don’t think a journal would want that? This excuse that there is too much at stake, too much credibility to be lost is weak. Do you really expect Dendrochronologists to go there and find out why they are wrong? Scientists need to be demonstrated to be wrong in the literature.

    Your argument seems to hinge on a huge cover-up. That there is a vast network of scientists collaborating to hide some ‘truth’ or to overstate the threat, or work together to avoid being found out so they can keep getting funding. It sounds like a conspiracy theory… Possibly the biggest in history.

    Do you know how much money Briffa got in funding to his study?

    “Consider that at Copenhagen world leaders are seriously considering the expenditure of trillions of dollars to control CO2 emissions based almost solely on the proposition that the current warming phase is “unprecedented” and is directly due to human activity. If the warming is not unprecedented and there are questions abouut the CO2 forcing built into the GCMs then it may be that the “Emperor has no clothes”, but many people are scared witless that this may be true and are fearful of the outcome to their reputations.”
    Well, do you know what it means if the MWP was as warm or warmer than now? What does that mean in terms of AGW. My understanding is that it implies greater climate sensitivity, the climate being less able to dampen pertubations.

  458. “If Steve thinks he has shown statistically that Dendrochronogy is seriously flawed, then he actually owes it to science to write that up and publish it. ”

    He doesn’t “owe” Science anything. He contributes what he chooses to. If his contributions are of value, then thats just the way it is, and you should stop whining about it. 😉

    Andrew

  459. Nathan: “Well, do you know what it means if the MWP was as warm or warmer than now? What does that mean in terms of AGW. My understanding is that it implies greater climate sensitivity, the climate being less able to dampen pertubations.”

    Actually Nathan it means that natural variability is greater than currently thought (by some) and that the current warming is of far less concern.

  460. Peter… ummm no.

    “NAthan, I’ve said on many occasions in the past that, if you are right and the Stick being wrong means that climate is more sensitive, then we should know and govern ourselves accordingly. And we should not thank those people whose obstruction has delayed the identification of this error. And I should not be the only person asking this question.

    Please continue this discussion at Unthreaded as it has nothing to do with Yamal.”

    That’s what Steve McI had to say about it.

  461. AndrewKY

    So you’d prefer he just kept it to himself, on his blog, and let us all suffer from not being blessed with his wisdom?

    Check out what he says about it:

    “…if you are right and the Stick being wrong means that climate is more sensitive, then we should know and govern ourselves accordingly. And we should not thank those people whose obstruction has delayed the identification of this error. And I should not be the only person asking this question.”

    Clearly he sees it as a ‘concern’. Shouldn’t he do the world a favour and help us understand?

  462. “So you’d prefer he just kept it to himself, on his blog, and let us all suffer from not being blessed with his wisdom?”

    No, the opposite. He makes a valuable contribution. He should keep contributing. But that’s his choice, not mine.

    You should follow his lead and choose to contibute something to this discussion besides logical fallacies. Something valuable.

    Andrew

  463. Nathan,

    Anyway your quote from him starts with:

    “if”

    If I was taller, I could be a basketball player. 😉

    Do we need to go over what “if” means?

    Andrew

  464. Nathan:

    As I have said here many times I am an editor. I do technical editing of geological work. I am very aware of the problems of getting things published, and with how and why people publish.

    This really is a pretty limited experiential basis to use for arguing how science communication works, and it still seems to be a bit arrogant on your part to come on this thread and try and explain to us the process by which scientific discovery gets made and communicated, given that limited background.

    You may be aware that one of the problems with the peer-review process is it can be exceedingly slow, and that in many fast-past fields of research, the rate of discovery often far outpaces the rate at which publication can occur.

    I’ve rarely seen an important question of our day that was decided via exchange of peer-reviewed articles. Traditionally the form of rapid communication was via exchange of letters between scientists (before modern transportation), or even letters from scientists to lay publications, and later via exchanges of views at important conferences.

    Probably the most famous of these was the 1927 Solvay Conference, in which the details and notation of classical quantum mechanics were hashed out.

    We have seen a rapid growth in the last five years of analytic blogs, and these have by and large accelerated the rate at which knowledge is circulated.

    Yes I agree that peer review is important, but if you are forced to resort to that as your primary means of learning about new results, you are probably going to be 2-years or so behind the research cycle.

  465. Andrew: Do we need to go over what “if” means?

    Right after we resolve what “is” means. 😛

  466. Carrick

    You continue to call me arrogant. Why?

    Anyway, you may be enlightened to know that I am aware of an exchange, using peer reviewed literature to resolve an ‘issue’. A woman at my work is helping put together a paper on stromatolites in the Amadeus Basin (in Western Australia and the Northern Territory) that will demonstrate that it is possible to use stromatolites as stratigraphic markers (ie use them to date rocks). I was speaking to her on the phone and she was expressing her joy at being able to finally prove to those “eggheads at Stanford” (at least I think it was Stanford) that stromatolites can be used as stratigraphic markers. Now apparently this is rare? I don;t think so. When scientists are in agreement they exchange via email, etc. Where people have big disagreements (and they are acually pretty common) the only way to ‘duke it out’ is via the literature, people are rarely on speaking terms.

    “Yes I agree that peer review is important, but if you are forced to resort to that as your primary means of learning about new results, you are probably going to be 2-years or so behind the research cycle.”
    I agree with you on this, but where you are trying to demonstrate someone is wrong you need the publishing power – it may take a few years but I am pretty sure Kath was enjoying every moment of her victory.

  468. You guys are ruining a very nice technical thread by feeding Nathan’s egocentric illness. Well done.

  469. “Just go to his blog and read it… Crikey it’s not hard mate!”

    I do. Everyday. I have been reading it everyday for a couple of years.

    Andrew

  470. lucia (Comment#22183)
    October 20th, 2009 at 9:54 pm
    Nathan/Carrick–
    “You are now arguing about the arguments in the thread instead of substance. I think it’s time to stop.”

    IMO everything after this comment should be zambonied.
    (This message will self-destruct in 10 minutes.)

  472. I understand. It’s too bad these knobs have to ruin a great thread. I wish people would tie their comments back to the OP. That’s the only way that blogs are going to generate anything of value.

  473. Nathan (Comment#22158) October 20th, 2009 at 5:42 pm
    Steven Mosher
    “If blogs science is so good, why does CA focus only on Published work? ”

    For the most part I don’t think that blogs generally “do science”
    There are some notable exceptions, posts here and there. Blogs, like journals, publish words. Sometimes they publish data and methods ( code). That in my mind amounts to “doing science”
    The reason, if in fact anyone needs to justify their action to you, that McIntyre focuses on Published science is varied. First off, I spent a long dinner with Steve where this topic came up. I’ll suggest that you do what I did. Invite him to dinner and probe his mind.

    “Why don;t they look at other ‘good blog science’ sites.”

    Hmm, well on ocassion ( you go count) posts at other blogs are discussed. Steve to his credit has a very narrow focus. As a marketing expert I would endorse this approach. As a life coach I would endorse this approach. As a Phd director I would endorse this approach.Personally, I focus my attention on the global temperature index. If I’m going to look at data in detail and look at code I better damn well focus.

    “As to your 2+2 = 4 claim. This is simplistic nonsense, the mathematics involved in proxy reconstructions is not trivial. ”

    I simplified it so you could get the point. Nature, for example, will not accept a paper that illustrates the math errors in a previous paper. That might get you a comment ( LT. 250 words)
    On the other hand SOME of the math in reconstructions is TRIVIAL. like core counts. Yamal=5.

    “If you can demonstrate that proxy reconstructions are flawed why do you hide it at your blog? Why wouldn’t you publish it? I think I know why, and that’s because there’s nothing really to publish. ”

    Well then you would be wrong. First, it’s not hidden at a blog. It’s published at a blog. So recently Steve Mc commented on a paper written by Rob Wilson. he pointed out an error. Wilson showed up. Made comments, noted the error. Tried to explain it. Now, what is the best way to fix this error? Submit a paper to journal?
    that takes a long time. A comment? equally long, and some journals cut off error corrections after 90 days.. kinda hard to catch errors without data and code.. hmmm. Your question is really Why publish on a blog rather than a journal.

    1. You can get the information out more quickly to more people.
    2. Your cost is lower.
    3. You avoid politics.
    4. more eyes = more reviewers = higher probability of finding the truth.
    5. no word limit
    6. no figures limit
    7. No page fees to the author.

    the real question is why would you publish in a journal!
    to get peer review? Why settle for anonymous review of 2 or 3 people when you can have thousands review it.

    “The ‘problems’ are trivial. You gave your JHanson example, why didn’t you add those data back in and quantify the difference? ”

    1. that wasnt the issue. the issue wasnt the effect of THIS deletion. the issue was the criteria and the method.
    2. The code as released required an AIX Box. recently, however, one guy has gotten the code to compile and run on a linux box.

    “Why did you just stop at that point and laugh? Why don;t you follow it through to it’s logical end point.”

    1. The logical point I wanted to make was simple. gavin claimed that the paper was COMPLETE. I said it wasnt and point out this one example. So, it’s not a matter of laughing. It’s a matter of the difference between an engineers idea of complete and gavins idea of complete. The end effect in this particular case doesnt matter. What matters is a sloppy process that may cause problems–like the Y2K error that McIntyre found. Like some of the night lights problems I found. Like the errors that EM Smith is finding as we speak. The point is if you are going to report a global index to a 1/100 C ( bogus in my mind) you had better cross your i and dot your ts ( get it)

    “The longer you postpone publishing your results, the longer you postpone quantifying your results, the longer you avoid following your work, the less scientists will take notice. Heck there isn;t even a summary page for all th published work CA has demolished. ”

    1. hmm. Well, the facts are actually against you on this one. i would say that more scientists are taking notice than at the start.
    I’ll just use an example from the global index fiasco. After WUWT did its survey, volunteer survey, did Noaa actually shut down some substandard weather stations? yup. As for CA, you need to look at some of the background of the millenium study.

    “If Steve won;t publish he could at least make up a summary outlining exactly the problem he has with various works. At it stands the site is a dogs breakfast. Look at the current post there on Yamal substitution. Are you saying he couldn’t publish a paper using that? If he outlined his argument using Polar Urals and published it then the dendochronologists would actually have something to debate.”

    You got time on your hands. have at it. Personally I have no trouble following steve, but I read it everyday. WRT a paper on the yamal substitution. You cannot get a paper out of it because of this. The WHOLE POINT is that there is no standard selection criteria for series. the dendros know this. There are “guidelines” and “recommendations” and post hoc screening. Its this fricking simeple. lets take RCS. How many cores do you need to do a valid RCS? the best I could find was a paper done on simulated tree cores which used 15 cores as a minimum But refused to endorse that as a minimum number required. or take EPS. In some publications people suggest using .85. in others ( where the series fall below .85 but above .8 ) they report that thet screened at .8. In yet another paper they screen at .7. That is witchcraft.

    ““I will make this simple. Until they release that data and the code they are not “doing science” They are publishing words and pictures.”
    This is so disingenous, you know that there are a lot of issues at play here. This is a cheap excuse for argument. What happened after GISS released it’s code? ” Write any papers on what is wrong with it? Anything happen? Or did you just discuss on your blog how awful it is without doing anything? Did you find the hidden ‘make it warmer smoking gun’?”

    I will differ to cheifio. But yes errors have been found. Those errors have been reported to GISS. Some have been fixed. Other have been recently identified we will see what the results are.
    You probably miss the attention to detail those of us who look at these things like. Finding an error of .02C is a big deal. for us.
    In the end once GISS is fully verified and tested. Then you would actually be in a position to begin to study whether it induced warming or cooling. But it’s 10K lines of undocumented code being looked at for the first time by an outsider. patience. My expectation? As I said to gavin over a year ago I don’t expect there to be ANY smoking gun that removes or adds SUBSTANTIAL warming. Don’t you read my posts from back in the day?

    What does being a Lukewarmer mean anyway?
    Basic tenets.
    1. Radiative physics is sound science. Adding GHGs warms
    the Planet.
    2. The Global temperature Index is most likely infected
    with some UHI ( at the limit .3C in the last century) but
    personally I would estimate .15C. we will see once all
    the data and code is released ( NOAA is still balking)
    3. GCMs ( as a whole) look to be biased high. Selected models
    perform better, BUT the community refuses to cull the herd
    and uses both good models and bad models to make projections. I brought this up to gavin and he replied that the community was not of a consensus on this, with some wanting to retain all models and others wanting to cull the herd.
    4. the MWP is a big question mark.

    So basically its getting warmer, GHGs are the likely cause, and alarmism is not the answer.

    “I’ve not seen any published work about luke-warming, I;ve not seen any analysis. ”

    sure you have. you just dont recognize it. Ross’s work on UHI.
    any work on mismatches between models and data. estimates of climate sensitivity. they just dont LABEL IT as luke warmer. we do. that’s called marketing.

    “Surely if you belive in luke-warming you have some sort of sensitivty estimate or forcing value for CO2 or ANYTHING AT ALL that you can publish about your theory. Where do you get your luke-warmer beliefs from?”

    1. to get a sensitivity estimate you would have to:
    a. correct the global index
    b. run a GCM
    2. I have no desire to publish anything in a journal. I believe that science journals should die. I’ve said so on many occasions.

    3. Where do I get my beliefs.
    a. reading and auditing the papers on UHI
    b. Analysis of temperature data.

    So basically, if you believe in GCMs ( I think they do a fair job given the right inputs– even said so on CA when I first figured out how to download GCM data for ModelE) and if you believe that the global index is skewed high since the 70s, then you
    have the fundamentals of lukewarming.

  475. Finding an error of .02C is a big deal. for us.
    In the end once GISS is fully verified and tested. Then you would actually be in a position to begin to study whether it induced warming or cooling.

    That’s an interesting position, Steven. Excuse me if I’ve misunderstood, but are you really arguing that one cannot be in a position to begin to analyse until one has observational certainty to the level you imply? Btw, I really don’t know what you mean by GISS ‘inducing’ warming or cooling…? Do you take the same view of HADCRUT and of satellite data? Will you be letting us know when you think we are in a position to begin to have the faintest idea of what the global temperature has been doing?

  476. Nathan, I can think of a few scientific disputes that were settled by peer reviewed publications, but not many.

    I agree that people who have really emotionally violent confrontations don’t generally communicate with each other via emails, but for a lot of those, there is a “showdown” that happens at a major conference or workshop, and often there is a feeling of resolution with one side a clear winner and the other a clear loser afterwards.

    Generally, even if you’re right and you publish, you can still expect people from the losing side of an argument get up while you’re talking and harangue at you a good ten years after the issue was “settled”. I know that for a fact, it’s happened to me. What usually happens is, because they push a dead theory, they no longer get published or funded, and if anything that situation just increases the ire on the part of the losing side even more.

    (I seriously think we need to retire this subthread now, if you don’t mind… I’ll give you last word if you want it.)

  477. Simon Evans,

    Will you be letting us know when you think we are in a position to begin to have the faintest idea of what the global temperature has been doing?

    Do you have the faintest idea of what global temperature has really been doing over the years?

  478. Dave Andrews,

    Yes. And we would have even if we had no thermometers. This is a bit O/T really, but it’s Steven who’s introduced the matter here. Perhaps we should let it lie for a more relevant thread – I was just really puzzled by some of his statements.

  479. “Yes. And we would have even if we had no thermometers.”

    So we can take or leave Lucia’s thermometer-derived Squiggly Lines? NO WAY. 😉

    Andrew

  480. Andrew_KY,

    Well,if Steven Mosher is right, then Lucia is simply wasting her time. Do you think so? I don’t :-).

  481. Simon,

    I do think Lucia has a misplaced belief in the Squiggly Lines, but I do not think she is wasting her time here. There is much to be explored and communicated about climate science and weather and related topics than just the politicized religion of AGW.

    Andrew

  482. Steven Mosher
    Too much to reply to.

    “Basic tenets.
    1. Radiative physics is sound science. Adding GHGs warms
    the Planet.
    2. The Global temperature Index is most likely infected
    with some UHI ( at the limit .3C in the last century) but
    personally I would estimate .15C. we will see once all
    the data and code is released ( NOAA is still balking)
    3. GCMs ( as a whole) look to be biased high. Selected models
    perform better, BUT the community refuses to cull the herd
    and uses both good models and bad models to make projections. I brought this up to gavin and he replied that the community was not of a consensus on this, with some wanting to retain all models and others wanting to cull the herd.
    4. the MWP is a big question mark.

    So basically its getting warmer, GHGs are the likely cause, and alarmism is not the answer.”

    But this is superficial nonsense. You think UHI is about 0.15C? Based on what? Where is your analysis?
    “we will see once all the data and code is released” – so it’s a guess?

    “the MWP is a big question mark. ”
    Why? I have asked many times what this means? WHAT DOES IT MEAN? Where is your analysis of what it means for a warmer MWP?

    “alarmism is not the answer”
    what has been done, so far, in all seriousness, that is alarmist? Nothing has been done. No steps have been taken.

  483. Steven Mosher

    I’ll comment on these 7 points:

    1. You can get the information out more quickly to more people.
    True

    2. Your cost is lower.
    True

    3. You avoid politics.
    I think you increase political debate. I think this is largely the role of blogs.

    4. more eyes = more reviewers = higher probability of finding the truth.
    Not true, look at WUWT. Many eyes, lots of garbage.

    5. no word limit
    This is a good thing? Being about to ramble on?

    6. no figures limit
    Again, is this a bad thing? Really? Being skilled in being succinct and to the point?

    7. No page fees to the author
    True

    Surely Steve could get his work published in a Statistics Journal.

  484. Bender

    “You guys are ruining a very nice technical thread by feeding Nathan’s egocentric illness. Well done.”

    What is wrong with you? You have some sort of Alpha Male thing going on B-boy.

  485. Ok, I’ll go. Bye, have a nice day.

    But please, at least do a post on what a warmer MWP means, I won’t post at it I promise… 🙂

  486. I just have to correct what I said above, I don;t know whether Steve thinks the MWP leads to higher sensitivity. I misread his quote.

    In my defence I did quote him in entirety. And I will again:
    “NAthan, I’ve said on many occasions in the past that, if you are right and the Stick being wrong means that climate is more sensitive, then we should know and govern ourselves accordingly. And we should not thank those people whose obstruction has delayed the identification of this error. And I should not be the only person asking this question.

    Please continue this discussion at Unthreaded as it has nothing to do with Yamal.”

  489. Nathan,

    “So what do you think a warmer MWP means?”

    That natural variability covers it easily and that everyone will be just fine like always.

    Now shut up and go back to bed!!

  490. Nathan,

    As far as I can tell there are 2 schools of thought on the climate system.

    The first school is ‘control freak school’ which presumes that the climate would remain constant unless acted on by some external forcing. This school also believes that climate sensitivity is a constant and is the same for all time scales and for all external forcings. For people that adher to this school a warm MWP must mean high climate senstivity because there are no large external forcings that can be associated with that period.

    The second school is the ‘shit happens school’ which presumes that the climate is a non-linear choatic system that could change states at any time for no particular reason. This school does not believe that climate sensitivity is necessarily a constant nor is it necessarily the same for all forcings. For people that adher to this school a warm MWP is evidence that internal, unforced variations in climate can be equal to or larger than the current observed warming. This means that the effect of CO2 is likely over stated by the GCMs because the GCMs do not take into account this internal variability.

    IOW, establishing that the MWP will resolve nothing. However, the issues with accountability, bias and access to data which the hockey stick arguments have undercovered are a sign of a systematic problem with science as it is practiced today. If we really want to get to a world where policies can be driven by “scientific evidence” then these problems must be addressesed.

  491. Raven,

    Why would the ‘shit happens’ school not use that position to say – for example – that AGW could well be much worse than we think, because rather than be in a ‘warming shit happens’ period we might equally be in a ‘cooling shit happens’ period, which is being completely overwhelmed by CO2 caused warming?

  492. David,

    The argument for aggressive CO2 mitigation policies is based entirely on the premise that climate predictable that it is possible to prevent certain outcomes if we only act in certain ways now. If climate is choatic and not predictable then it is impossible to do any cost benefit analysis. This means we will be forced to do what we do with other natural disasters like earthquakes and hurricaines.

  493. Raven

    I raised the issue because RyanO had claimed that there was an attempt amongst scientists who do proxy reconstructions to try and ‘flatten’ the MWP. I was trying to work out if that is actually a sensible suggestion – and it seems it isn’t.
    There is a possibility that the aim was to make the current warming seem more important, as some others suggest, but why pretend about something that actually isn’t important either way. It is a pretty weird sort of aim.

    What I think more likely is the scientists make these claims, because they think they’re right and justified, and people like yourself (and others here and at CA) see their work as not justified. Which then leads to the question of how do you resolve that? I can only see resolution in a few ways:

    make reconstructions that are better and more convincing

    write a general refutation of the methodology, perhaps as Lucia tried here, and maybe submit to a stats journal

    Attempt to work with them to ‘fix’ the problem (if it exists).

    Can you see any other way through?

  494. Nathan:

    I raised the issue because RyanO had claimed that there was an attempt amongst scientists who do proxy reconstructions to try and ‘flatten’ the MWP. I was trying to work out if that is actually a sensible suggestion – and it seems it isn’t.

    The reason that it is desirable, if they are “alarmists”, to not have a dominant MWP, is because they can then claim that this current warming period is “unprecedented”, which strengthens the argument for “acting now”.

    So, yeah, it’s a sensible suggestion and it’s your logic that “isn’t”.

  495. Michael Mann’s own words (via the Wikipedia):

    It is the consensus of the climate research community that the anomalous warmth of the late 20th century cannot be explained by natural factors, but instead indicates significant anthropogenic, that is human influences… More than a dozen independent research groups have now reconstructed the average temperature of the northern hemisphere in past centuries… The proxy reconstructions, taking into account these uncertainties, indicate that the warming of the northern hemisphere during the late 20th century… is unprecedented over at least the past millennium and it now appears based on peer-reviewed research, probably the past two millennia.

  496. Raven

    “The argument for aggressive CO2 mitigation policies is based entirely on the premise that climate predictable that it is possible to prevent certain outcomes if we only act in certain ways now. If climate is choatic and not predictable then it is impossible to do any cost benefit analysis. This means we will be forced to do what we do with other natural disasters like earthquakes and hurricaines.”

    Do we have any good evidence that the climate is that unpredictable? Do we see such wild swings (ones that apparently have no cause; that would make mitigation useless) at any time in the Holocene? Or even at any time in the Phanerozoic? The modelling on the climate during the Phanerozoic seems to be pretty good with geological evidence of climate.

    http://www.gsajournals.org/archive/1052-5173/14/3/pdf/i1052-5173-14-3-4.pdf

    That’s been kicking around for a while.

  497. Carrick

    So they are engaged in a conspiracy to hide the truth warmth of the MWP, so they can claim that the current warming is unprecedented. Even though in terms of modelling and projecting into the future it doesn’t really matter – in fact the future could be worse… It’s not an argument they needed to present. It’s a pretty lame conspiracy theory, since the argument for action still remains even without the MWP. There is no need for them to hide the warmth of the MWP.

    Also
    “unprecedented over at least the past millennium and it now appears based on peer-reviewed research, probably the past two millennia. ”

    If the MWP was sometime between 900AD and 1300AD and peaked around 1000AD, he could still be correct even if the MWP was as warm as today. It’s still unprecendented over the last 700 years…

  498. Given that CO2 generally (but not always) lags temperature by about 800 years, it appears that the CO2 levels are responding to temperature changes rather than driving them.

    Also, you’re reference (which appears badly flawed in its premise due to causality problems, rather a noob science error don’t you think?) has nothing to do with whether global climate models can effectively predict climate changes.

  499. Nathan:

    So they are engaged in a conspiracy to hide the truth warmth of the MWP, so they can claim that the current warming is unprecedented

    What gives with you?

    Nobody said anything about a conspiracy, just bad science.

    The rest of your comment is based on this flawed premise, and I’ll ignore it.

  500. Nathan:

    It’s still unprecendented over the last 700 years…

    Ever hear of the Little Ice Age?

    We don’t need anything as radical as dendro climate proxies to reconstruct that the Little Ice Age existed, and that it was colder during that period. “Duh.”

  501. Carrick

    Ok, get rid of the word conspiracy and add the word accidentally.

    “Given that CO2 generally (but not always) lags temperature by about 800 years, it appears that the CO2 levels are responding to temperature changes rather than driving them.

    Also, you’re reference (which appears badly flawed in its premise due to causality problems, rather a noob science error don’t you think?) has nothing to do with whether global climate models can effectively predict climate changes.”

    Ummmm are you serious? I mean this is pretty basic stuff…

  502. Actually Carrick

    After your comment about causality etc I don’t think we have much common ground so best if we use your suggestion…

    I’ll also say “The rest of your comment is based on this flawed premise, and I’ll ignore it. ”

    So we can now safely ignore each other.

  503. Nathan, it is also very hard to explain a MWP with the natural factors that the framework of AGW allows into the equation.

  504. Nathan:

    Ok, get rid of the word conspiracy and add the word accidentally.

    That actual doesn’t help your argument. Try reframing it.

    Also are you seriously suggesting that CO2 lagging temperature changes by an average of 800 years is consistent with CO2 being a causative agent?

  505. Carrick,

    The reason that it is desirable, if they are “alarmists”, to not have a dominant MWP, is because they can then claim that this current warming period is “unprecedented”, which strengthens the argument for “acting now”.
    So, yeah, it’s a sensible suggestion and it’s your logic that “isn’t”.

    So, by your logic, then, before the publication of MBH 98/99 ‘they’ would not have been making an “unprecedented” claim and a determination to act now?

    The trouble is that such ‘logic’ doesn’t stand up to examination. From the IPCC’s First Assessment Report, published eight years earlier, we had:

    …emissions resulting from human activities are substantially increasing the atmospheric concentrations of the greenhouse gases: CO2, methane, CFCs and nitrous oxide. These increases will enhance the greenhouse effect, resulting on average in an additional warming of the Earth’s surface….

    … We calculate with confidence that: …CO2 has been responsible for over half the enhanced greenhouse effect; long-lived gases would require immediate reductions in emissions from human activities of over 60% to stabilise their concentrations at today’s levels…

    …Based on current models, we predict: under [BAU] increase of global mean temperature during the [21st] century of about 0.3 oC per decade (with an uncertainty range of 0.2 to 0.5 oC per decade); this is greater than that seen over the past 10,000 years…

    – and that was the report, of course, whose only graph of indicated NH temperatures was taken from Lamb’s Central England Temperature reconstruction, which displayed an MWP equal to the CWP of the time.

    It is thus demonstrably clear that the assessment of the global CWP being unprecedented and of the need to act now pre-dated the MBH papers.

  506. Carrick

    “Also are you seriously suggesting that CO2 lagging temperature changes by an average of 800 years is consistent with CO2 being a causative agent?”

    Yes. becuause there is more than one cause, We know the Milankovitch Cycles have been driving the ice ages and CO2 has moved in response to that. The Orbital factors drove the initial response, then CO2 moved and assisted. Looking at the present day we have lifted CO2 levels – which should then warm the Earth despite the Orbital factors not changing… Well that’s my simplified version.
    Seriously you should find some literature on this because it’s been done to death.

  507. Niels

    “Nathan, it is also very hard to explain a MWP with the natural factors that the framework of AGW allows into the equation.”

    Yes, so given our current understanding it’s doubtful the MWP was warmer than today. UNLESS you invent some imaginary factor that allows it to be warm. I am not saying such a factor doesn’t exists, but it’s not a sensible proposition. Becuase if there is some unknown factor that made it warmer (that is in play now), why is there not some other unknown factor that cancelsthat factor out, and perhaps there is an infinite array of imaginary factors that cancel each other all the way down… It’s a bit of a rabbit hole.

  508. Also are you seriously suggesting that CO2 lagging temperature changes by an average of 800 years is consistent with CO2 being a causative agent?

    Yes. The climate is a complex system, and there are several forcings that can be the most powerful at any one point in time due to the environment that exists at that time. Also consider, we have not been around for long in the geological record. Looking for an anthropogenic factor in past climate isn’t going to get you far.

  509. Simon, several problems there.

    First Nathan was asking what motivation would there be to “attempt […] to try and ‘flatten’ the MWP”. Claiming the warming is unprecedented is one of those, and that observation does not hinge the notion that the “unprecedented” claim was given birth to by MBH 98/99. I never said or suggested such a thing.

    Secondly… in terms of the “unprecedented warming”, it is now generally understood from GCMs that anthropogenic effects (CO2 and sulfate emissions principally) more or less balanced each other until the 1970s… see for example this. Note that SO2 and CO2 emissions more or less balanced each other until CO2 emissions went into “overdrive” in the ’70s.

    If we look at the 20th century there are two well-established periods of warming, circa 1910-1940, where we say a 0.45°C temperature change and 1970-2005, where we say a 0.55°C temperature change.

    The first of these “unprecedented” temperature shifts, according to the GCMs, couldn’t be due to man, so a) either there is a fluke and we had a natural “unprecedented” temperature shift followed by a man-man one just 30 years later, or more likely b) there is a problem with the claim of “unprecedented”.

    I’ll note the interest in pushing a particular political agenda predated MBH 98/99.

  510. Nathan:

    Yes. becuause there is more than one cause, We know the Milankovitch Cycles have been driving the ice ages and CO2 has moved in response to that.

    Since you keep losing track of your own arguments, I’ll remind you that you posted a link to an article entitled “CO2 as a primary driver of Phanerozoic climate”.

    Basically what they showed is that CO2 and temperature correlate with each other. I noted that CO2 lags temperature, so we know that CO2 can’t be what drives climate change. (I never stated it couldn’t amplify it, you seem to have a lot reading comprehension issues for an editor.)

    Anyway, we’re back to the “correlation is not causation” statement here. Since CO2 lags temperature, rather than the other way around, we can state that CO2 is not the prime cause of the temperature shifts.

    So what are we left with?

    Not so much, because this article doesn’t model the climate, it simply points out that a phenomenological relationship between temperature and CO2 exists. Yet you can get this effect even if CO2 did not drive temperature at all (I don’t endorse that view btw) simply by invoking a model in which other natural climate forcings were larger, and CO2 was simply going into or coming out of solution in the oceans, in response to temperature changes of the oceans in response to these other natural climate forcings.

    The data are also consistent with the stated CO2 atmospheric concentrations changing in response to external forcings, and then amplifying them by as much as 87%.

    So think about it, we have a range 0 to 87% amplification factor that is allowed by the data. Any number in this range is consistent with the data in your linked publication.

    So your publication proves nothing.

    Seriously you should find some literature on this because it’s been done to death.

    Oh what piffle. Stick to facts, and if you can’t argue them, admit it.

  511. Carrick

    How could you possibly see an 800 year lag in a 500 million year reconstruction. That is simply stupid.

    I’m nt going to debate your points because they are too stupid. If you don;t think CO2 is a climate forcing, good luck to you.

  513. Carrick

    How can you SEE an 800 year lag in the 500 Million year long Phnerozoic? The resolution isn’t clear enough. We know there is an 800 year lag for the last few ice ages, and it is well explained.

    We DON’T KNOW if it lags by that much throughout the Phanerozoic. Nor is there any particular reason to asume that. The explanation for why it lags implies it lags specifically in the Holocene, but not much else. Why are you assuming it lags throughout the Phanerozoic?

    “Anyway, we’re back to the “correlation is not causation” statement here. Since CO2 lags temperature, rather than the other way around, we can state that CO2 is not the prime cause of the temperature shifts.”
    It doesn’t need to prove the causation, that’s an assumption. We know that CO2 has certain properties… Have you heard of the Greenhouse Effect? If the CO2 is there, we know the Earth will be hotter than without it.

  514. Nathan (Comment#22269)
    October 22nd, 2009 at 2:11 am

    “I raised the issue because RyanO had claimed that there was an attempt amongst scientists who do proxy reconstructions to try and ‘flatten’ the MWP. I was trying to work out if that is actually a sensible suggestion”

    First, you are aware of the quote that proves at least one person has thought it would be a good idea to generate unprecedentedness by manipulating the data to lower the MWP?

    Second, did you read Broecker (2001) and the reply to it from Bradley, Mann & Hughes?

    The fight over the magnitude of MWP has nothing to do with climate sensitivity. How many proofs and explanations does a guy need?

  515. Do we see such wild swings (ones that apparently have no cause; that would make mitigation useless) at any time in the Holocene?

    We see them in some of the climate models. . . 🙂

  516. Carrick,

    What, in your view, is the best evidence of a ‘dominant MWP’ occurring concurrently on a global scale? I don’t see how one can suggest ‘flattening’ an MWP without first establishing that it is ‘unflattened’.

    On your second point:

    it is now generally understood from GCMs that anthropogenic effects (CO2 and sulfate emissions principally) more or less balanced each other until the 1970s… see for example this. Note that SO2 and CO2 emissions more or less balanced each other until CO2 emissions went into “overdrive” in the ’70s.

    Hmm – “more or less”? Your link shows GHG against all other forcings, btw. But anyway, where has a claim ever been made that the temperature change 1910-1940 was unprecedented? Most palaeo reconstructions would suggest that it wasn’t (Mann et al 2008, for example). The unprecedented assessment is made on a centennial scale, and considers the extent of the current anomaly rather than variance over a thirty year period.

  518. bender,

    First, you are aware of the quote that proves at least one person has thought it would be a good idea to generate unprecedentedness by manipulating the data to lower the MWP?

    Do you mean “we have to get rid of the MWP”?

    If so, I’m aware that David Deming has made the claim he received an email containing those words, however I have seen no proof of the matter. Can you direct me to such proof, if it exists?

  519. “Do we see such wild swings at any time in the Holocene?”
    The Holocene is nothing but “wild swings”.

  520. bender,

    I posted before seeing your link to CA. That is, of course, not proof. Has there ever been any evidence presented to verify Dening’s claim?

  521. Simon Evans:
    I’ve never seen any such proof. AFAIK it could be a fabrication by Deming. I don’t know that Overpeck has ever been asked the question, or if he was, what his response was.
    .
    What is more telling – because it’s all on the record – is the exchange between Broecker and The Team in the 2001 Science article “Was the medieval warm period global?”.

  522. Is it a coincidence – Deming’s assertion – and the substance of, and motivation behind, the Broecker exchange? Possibly.

  523. The Milankovitch Cycles don’t really match up with the ice age cycles. The Earth actually gets more annual solar irradiance just as the ice age cycles are starting to really cool down. There are changes at high latitudes that match up closer but high latitude solar irradiance cycles up and down at about 25,000 years periodicity compared to the 100,000 year ice age cyle.

    And CO2 only provides for a partial explanation of the temperature change during the ice ages (albeit at a lag of 800 to 2,000 years).

    http://img261.imageshack.us/img261/2127/last3iceages.png

    Royer and Berner’s CO2 and temperature numbers from “CO2 as a Primary Driver of Phanerozoic Temperature” only point to about 1.0C to 1.5C per doubling. Nobody bothered to do the math.

  524. Nathan’s underlying question seems to be this:
    ‘Why would anyone try to disappear the MWP if this would have no implications for climate sensitivity?’ He can’t see the connection, so he’s trying, first, to understand what might be wrong about either side of the equation. And second, to clarify any connection that might exist.
    .
    The answer – the connection – is that the alarmist call to action hinges on more than one assertion. First, that current temperatures are already alarmingly high; second, that they will continue to rise higher. The first is dependent upon the magnitude of the MWP. The second is dependent on the climate sensitivity calculation. The magnitude of MWP is not strongly connected to the climate sensitivity calculation*. The link between them is more political than scientific.
    .
    *But see my related argument at CA, along the lines of Raven’s bit on intrinsic natural variability.
    .
    It’s hard to understand what exactly Nathan’s after, but I think this is at the core of it.

  525. Bender…
    The “wild” swings in the Holocene are pretty nicely explained by the Milankovitch Cycles. Not some mysterious unknown factor.

    Well you have some text… Don’t know if that’s a quote. I will look at that other text you wrote.

    In any case, what is the “need” for a warm MWP?
    The only case I see you have is that a cooler MWP Is important for selling AGW. Ummmmmm that’s not very convincing. And you have Loehle’s reconstruction that says it as warmer yes? Hmmmm that didn’t seem to go vey well either.

    So. You think there is a group of scientists attempting to portray the MWP as cooler, as a selling point. You don’t know anything in particular about the MWP other than what is written in the Norse writings.

    But. There’s no particular need for the MWP to be cooler to prosecute a case to act on AGW. No modelling can recreate a warmer MWP (using the current understanding of climate) without invoking imaginary parameters. There’s no particular evidence around the world that it was much warmer. They’re pretty big problems. At least a cooler MWP doesn’t have to invoke imaginary factors. It’s always good when a branh of science is supported by an unrelated branch yes? So the physics of climate modelling, supports the evidence of the proxies. You may have problems with the proxies, I don;t know if they’re doing it right. But at least it is supported by our physical understanding of climate.

    And. We know that there WAS a warmer period, a bit earlier in the Holocene optimum (10000 BP). They don’t seem scared of that, why not? Because that’s nicely explained by Milankovitch.

  526. Basically what they showed is that CO2 and temperature correlate with each other. I noted that CO2 lags temperature, so we know that CO2 can’t be what drives climate change. (I never stated it couldn’t amplify it, you seem to have a lot reading comprehension issues for an editor.)

    It can, it’s basic physics, and Lucia and McIntyre seem to agree. They just disagree on the degree forcing.

  527. “But at least it is supported by our physical understanding of climate.”

    When the “understanding of the climate” is poor, this doesn’t really mean anything.

    Andrew

  528. “Lucia and McIntyre seem to agree”

    Again the Appeal to Authority. While I respect both, their names don’t confer infallibility on them.

    Andrew

  529. Bender
    What I want to know is do you know what it means if the MWP is warmer than now – in terms of the science of AGW. Does it mean that we can just stop and not take action. I think the answer to that is “no”.

    “The answer – the connection – is that the alarmist call to action hinges on more than one assertion.”
    Actually the call to action happened well before the Hockey Stick. Hansen has been making the call since the early 70s. Heck Arthur C Clarke was saying the 40s or 50s we should burn coal to avoid the next ice age.

    Well Broeckers paper was very nice… He doesn’t really get very specific. Was there something in it you wanted me to see specifically?
    “The case for a global Medieval Warm Period
    admittedly remains inconclusive. But keeping
    in mind that most proxies do not have adequate
    sensitivity, it is interesting that those capable
    of resolving temperature changes of less
    than 1 “C yield results consistent with a global
    Medieval Warm Period.”
    That’s the opening of his conclusion.

    “I consider this evidence sufficiently
    convincing to merit an intensification of
    studies aimed at elucidating Holocene climate
    fluctuations, upon which the warming
    due to greenhouse gases is superimposed.”
    That’s the end. So he says we need further work (this is around 2001)

    Did he do anymore work?

  530. For those just tuning in (or those who can’t seem to grasp on the Nth try): Science isn’t about personalities. It’s about evidence.

    Andrew

  531. bender,

    I’ve read the Broecker article and Bradley et al’s letter of response in Science. Is there further exchange beyond that?On the basis of those two items, I can’t see what point you are making about “motivation”. Bradley et al do not agree with Broecker that only boreholes and snowlines can reconstruct temperatures to within 0.5C, and they point out that selecting only a few data points “adds little” since it is known that regional variations occur against the global trend. In what way is this “telling”, other than telling us that they don’t think much of Broecker’ paper?

  532. Nathan,
    .
    “There’s no particular need for the MWP to be cooler to prosecute a case to act on AGW.”
    .
    “Acting” on AGW will cost extraordinary amounts of money and can only be justified if the expected return on investment exceeds the cost of not acting (a.k.a. adaptation). If climate is unpredictable then it is not possible to demonstrate that the “action” makes economic sense.
    .
    e.g. If we knew with 90% certainty that a dinosaur killer meteor would strike within 50 years then it would be possible to justify investing trillions in the space technology that could divert the meteor. The trouble is we can’t know that with any certainty even though it is a plausible event. This means all we can rationally do is invest a small amount money in trying to detect objects that might strike earth. This is the situation I feel we are in with climate.
    .
    “No modelling can recreate a warmer MWP (using the current understanding of climate) without invoking imaginary parameters.”
    .
    Only because the current understanding simply assumes that that climate cannot change without an external forcing which requires that “imaginary parameters” be invented. There is no need for such parameters if one treats climate as a chaotic non-stationary system.

  533. Did Broecker know about the Hockey Stick in 1987?

    http://muse.lib.nchu.edu.tw:9797/MuseSessionID=eadffbf0c6946c9a52a39876a78121e/MuseHost=www.nature.com/MusePath/nature/journal/v328/n6126/pdf/328123a0.pdf

    Or did he change his mind and stop being an alarmist later?

    This is pobably more your thing though:

    http://www.ncbi.nlm.nih.gov/pmc/articles/PMC34299/

    “One aspect of the debate with regard to the extent of global warming generated by the ongoing buildup of greenhouse gases has to do with the contribution of natural climate change. Those opposed to emission restrictions are quick to conclude that the warming during the past century may be dominated by the relaxation of cold LIA conditions. Although we are still a long way from being able to assess whether or not this interpretion is correct, were thermohaline circulation to be implicated, a step toward this goal would be taken.”

    So maybe you are being “too quick”?

    http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6VF0-4M6459T-1&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&_docanchor=&view=c&_searchStrId=1059550258&_rerunOrigin=scholar.google&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=827ba5e670038ecc7a9371a0b53be4b3

    That’s the latest piece I can’t find by him – but can’t read it.
    Are you sure he’s a “no action” kind of guy? You may be able to use his work to promote an alternate theory (that the MWP was global) but it doesn’t seem like he’s promoting no action

    http://en.wikipedia.org/wiki/Wallace_Smith_Broecker

    Heck he even coined the term “gobal warming” apparently.

  534. Raven

    Thus we come to the centre of the rabbit warren

    ““Acting” on AGW will cost extraordinary amounts of money and can only be justified if the expected return on investment exceeds the cost of not acting (a.k.a. adaptation). If climate is unpredictable then it is not possible to demonstrate that the “action” makes economic sense.”

    Yes. It’s just about money. Hmmmm so rather than these scientists attempting to flatten the MWP to cause alarm. Is it not more likely you that you want to enlarge it so as to promote the “we don;t need to do nuthin” Line. See that argument makes sense. Because a warmer MWP actually makes no difference to whether we take action.

  535. Raven

    “Only because the current understanding simply assumes that that climate cannot change without an external forcing which requires that “imaginary parameters” be invented. There is no need for such parameters if one treats climate as a chaotic non-stationary system.”

    This is very easy to say. Would be a lot more useful to your argument if you actually made that model and demonstrated it, yes?

    Your argument seems to be that becuase there are some uncertainties, we should do nothing. Well, frankly, that’s not very convincing.

  537. “actually made that model and demonstrated it, yes?”

    How do you model chaos, Nathan?

    Andrew

  539. Andrew, even if the system is stochastic, there are properties of the system whose evolution can be modeled over time. In the case of climate, one such parameter would be the long-term average of temperature.

    The question at hand here is not whether the system is stochastic so much as whether the natural fluctuations are large compared to the ones predicted from human activity.

    That’s why they invoke “unprecedented warming”, even though as I pointed out the “unprecedented warming” also involved what the GCMs tell is must be a large natural “unprecedented warming”.

  540. Nathan,
    .
    “This is very easy to say. Would be a lot more useful to your argument if you actually made that model and demonstrated it, yes?”
    .
    And how would you demonstrate that such a model was correct? If it correctly ‘predicted’ the future it would be a fluke. Our scientific institutions are not set up to deal with problems that cannot be analyzed because such systems do not provide a justification for funding. That is why climate scientists prefer to view climate as something that can be modelled.
    .
    “Your argument seems to be that becuase there are some uncertainties, we should do nothing. Well, frankly, that’s not very convincing.”
    .
    Then you should be in favour of spending trillions evacuating Los Angles because the science tells us there might be a massive earthquake any day. We do not have the resources to act on every plausible risk. We must make choices based on the likeyhood of an adverse outcome vs. the cost of acting to avoid it. If climate is unpredictable then we cannot justify the massive investments required to meet the specific targets demanded by environmentalists (e.g. 450ppm or 350ppm or whatever).
    .
    That said, there are many actions that could make sense even with the uncertainity. The trouble is we are not allowed to discuss those options as long as the ‘climate is predicatable’ crowd controls the adgenda and insists that policies be designed to meet specific targets within specific times. Targets that only make sense if one assumes that climate is predictable.

  541. Nathan,
    .
    “Yes. It’s just about money. Hmmmm so rather than these scientists attempting to flatten the MWP to cause alarm. Is it not more likely you that you want to enlarge it so as to promote the “we don;t need to do nuthin””
    .
    The anecdotal evidence for a warm MWP is quite strong which means the onus is on the people claiming that this anecdotal only represented a regional phenomena. The fact that Mann, Briffa and others have been forced to rely on carefully selected proxies processed with “novel” statistical methods suggests that the case for a cool MWP is extremely weak. If there case was strong they would not have had to resort to such tactics.

  542. “The question at hand here is not whether the system is stochastic so much as whether the natural fluctuations are large compared to the ones predicted from human activity.”

    I think that IS the question. Do you know if the *global* climate system is stochastic or not? How do you know what the answer is?

    Andrew

  543. as I pointed out the “unprecedented warming” also involved what the GCMs tell is must be a large natural “unprecedented warming”.

    No they do not! I repeat, nobody has claimed that the relative temperature change 1910-1940 was unprecedented and nor, for that matter, is the relative change in temperature 1970-2000. What is considered unprecedented is the current extent of anomalous warmth, along with the length of time the temperature has been at or above the assessment of a previous peak. The GCMs actually reproduce most of the 1910-1940 period quite readily through natural forcings only, and some runs will, of course, show temperature change in excess of the record for that period. What they will not do is reproduce the whole of the 20th century without anthropogenic forcings. That is what the GCMs tell, not that 1910-1940 change was ‘unprecedented’.

  544. Nathan,

    “‘ the call to action happened well before the Hockey Stick. Hansen has been making the call since the early 70s. ”

    Please provide a reference. The only things I have read were that Hansen provided code for another gentleman at his University in the EARLY 70’s. That other gentleman was working on the Ice Age scare!!!! While Hansen was working on Infrared measurement on Venus probes and similar, I think it would be a stretch that he would have enough information at that point to have done anything other than GUESS at Climate.

    If you move your claim to the late 70’s I would agree.

    In other words, if he WAS sounding the alarm at that time, it was with NO FACTS and would support denialist contention that he is an ACTIVIST with an AGENDA unsupported by evidence.

    I would remind you of the current requirement that we have 30 years data to establish trends. In the early 70’s what trend was there?? In the late 80’s when Hansen made his activist statement to Congress he only had a 10 year trend of dubious signifcance.

    For that matter, here we are 21 years later and the trend has turned. (yes, I am a visionary like Hansen and can predict this with absolute certainty!!!)

  545. Simon:

    I repeat, nobody has claimed that the relative temperature change 1910-1940 was unprecedented and nor, for that matter, is the relative change in temperature 1970-2000.

    Actually the warming trend 1910-1940 is almost statistically indistinguishable from the 1970-2000 warming period in terms of the total temperature change and the rate of increase in temperature. If you are going to say that one of these is in someway remarkable, you’re pretty much stuck saying the same about the other.

    Secondly, we only have a short period where good instrumentation coverage is available. And we know the first 50 years of that was an overlap from the end of the LIA, so the fact we have warmed from an ice-age like climate to a more temperate one than that doesn’t say very much.

    Thirdly, instrumentation has a much finer time resolution than is available with currently used proxies. If you use proxy measurements (so you are comparing apples to apples) you don’t find, except in a few anomalous specimens, that the 20th century is particularly extraordinary. (That isn’t the same thing as saying is it isn’t extraordinary, just that the time-resolution of the currently know proxies is too poor, given other correlated measurement noise sources to be able to resolve an event that short of a interval.)

    The GCMs actually reproduce most of the 1910-1940 period quite readily through natural forcings only, and some runs will, of course, show temperature change in excess of the record for that period

    That’s a pretty vague statement with regards to how well the GCMs reproduce the early 20th century warming, but this is a separate topic. I happen to disagree with you that the models do an adequate job of explaining that warming period, though I also believe part (but not all) of this almost certainly is due to instrumentation error.

    But if we stick to the question of evidence for unprecedented warming, then the answer is “not so much”.

  546. Raven:

    Then you should be in favour of spending trillions evacuating Los Angles because the science tells us there might be a massive earthquake any day. We do not have the resources to act on every plausible risk. We must make choices based on the likeyhood of an adverse outcome vs. the cost of acting to avoid it.

    More than that, we do have other pressing world needs, such as third-world poverty, AIDs, malaria, and a host of other problems that a modest amount of expenditure could result in millions of lives saved. The Copenhagen Consensus has an enumeration of the various problems facing our world, and typically the cost-versus-benefits of global warming amelioration compared to other needs ends up placing it at the bottom of the list.

  547. Carrick,

    Actually the warming trend 1910-1940 is almost statistically indistinguishable from the 1970-2000 warming period in terms of the total temperature change and the rate of increase in temperature. If you are going to say that one of these is in someway remarkable, you’re pretty much stuck saying the same about the other.

    Excuse me, but I was saying precisely the opposite of what you suggest, in the very words that you have quoted from my post, viz.:

    I repeat, nobody has claimed that the relative temperature change 1910-1940 was unprecedented and nor, for that matter, is the relative change in temperature 1970-2000.

    It is quite beyond me how you can quote that and then suggest that I am saying that either period of relative temperature change is in itself remarkable. May I remind you that you were earlier making snarky comments about another poster’s comprehension! Enough of this – there is no point in repeating myself when you have already shown so clearly that you don’t understand what you read.

  548. Simon, I apologize that I didn’t properly parse your text, and am not sure I understand what you are trying to drive at.

    So why are you saying “nobody has claimed that the relative temperature change 1910-1940 was unprecedented and nor, for that matter, is the relative change in temperature 1970-2000.”

    Mann’s “unprecedented warming” certainly hinges in the temperature shift from circa 1850 to current.

  549. Carrick,

    Well, that’s gracious, thank you, so I will respond again.

    1850 to current is very obviously a longer period than 1910-1940 or 1970-2000! If you were to imagine any two periods of natural temperature gain equivalent to either of those two shorter periods occuring back to back over a century then you would expect to have seen a natural change which was, in total, unprecedented in terms of relative temperature change. If such a natural change had taken temperatures to a level above those assessed to have been previous peaks then that, also, would have been unprecedented. All of this would be true of natural temperature change – the observation of whether or not it is unprecedented is regardless of attribution.

    Now, you or others may dispute whether or not the assessment of earlier temperature levels is reliable, but that is beside the point of what is meant by unprecedented. To the extent that the reconstructions are reliable, the temperature change over the whole period is unprecedented, regardless of whether it has been the outcome of natural forcings, anthropogenic forcings or the mischief of the fairies at the bottom of my garden.

  550. Raven,

    So those who think that climate is unpredictable suggest that we have – effectively – zero knowledge of the cause and effect relationships operating in climate?

    I think that that position would not be held by too many people. It seems to me that we have very good knowledge (but not complete knowledge) of many (but not all) drivers of climatic response. The MWP argument would seem to be, ‘We do not know the cause of this event; therefore, we do not know enough about climate to make any accurate predictions.’ If that was the attitude to the rest of science, we would have made zero progress in astronomy, cosmology, particle physics and so on and so on.

    Not knowing some things about the climate system does not equal not knowing enough about the climate system to make accurate predictions.

    In fact, the whole CO2/forcing issue demonstrates that we do indeed know enough about the climate to make accurate predictions – unless, of course, through sheer random chance the match between predicted and observed forcing due to CO2 was matched by some unknown forcing while at the same time CO2 turned out to be not a warming agent at all.

    I think that those who rely on this argument to set aside AGW are not on the right track.

  551. Raven

    If your main complaint is “TRIILIONS OF DOLLARS”. Welll that’s a pretty poor substitute for debate. I have learnt enough of why you think the MWP is important. And I think it’s pretty poor theory.

    Kuhnkat
    I know Hansen wrote some papers about the enhance greenhouse effect prior to the global cooling nonsense. I’ll find them and post them. The point I am making is that people understood that pumping CO2 into the atmosphere was a bad thing well before the Hockey Stick. The call to action came well before the Hockey Stick. There is no particular need for the Hockey Stick to exist to have a call to action.

  552. Well said David
    It would seem Raven would rather ignore all the knowledge we have because that knowledge isn’t perfect. It is the same sort of arguments used by the tobacco lobby and ID proponents. It goes like this: Because our knowledge is imperfect, we cannot be certain we know anything. It is a poor argument.

  553. Nathan (Comment#22269)
    .

    I raised the issue because RyanO had claimed that there was an attempt amongst scientists who do proxy reconstructions to try and ‘flatten’ the MWP.

    .
    You seriously need to stop ascribing things to people who said nothing of the sort. If you’re referring to my comment at CA in the Yamal thread, I actually said the opposite of the above with respect to Rob Wilson, to whom the comment was addressed. At this point in time, I make to you the following suggestions:
    .
    1. Read more [carefully].
    2. Post less.

  557. RyanO

    “My honest assessment of the community is that (generally speaking) it is deliberately trying to over-reach its ability to draw inferences about past climate by biased sampling and archiving, poor use of statistics, habitual refusal to disclose data and methods, and lack of rigorous procedures for selecting and processing data.

    Hmmmmmmmmmmmmmmmmmmmmmmmmmmmm

    Interesting. And yet here we have so many people who do seem to believe that… Bender. Raven. Kuhnkat. AndrewKY…

    Welll it’s welcome news that you DON’T think that, and I apologise if I misinterpreted what you wrote above.

    So if you don’t believe that, why is there so much passion about it? If you don’t think the community is trying to make the MWP flatter, then why all the angst? Why?

    “They are all important issues, since the output of this field has had – and continues to have – a dramatic political influence that will affect every one of us.”

    Well yes, because you see it affecting our political future…

    But… The proxy reconstructions don’t matter in terms of the call to action against global warming.

  558. Nathan (Comment#22338)
    .
    Right. So where in there did I say Wilson was trying to flatten the MWP? My complaint is not any “flattening” of the MWP – my complaint is shoddy science passed off as a near-certain result. It is the certainty with which dendros present their work that I take issue with.
    .
    I also find myself wondering why you did not copy my whole quote, as this point is quite clear in context:
    .

    “Thank you for dropping by (several times). I apologize for the bluntness of the following (especially since you seem to be one of the good guys), but since you dropped by instead of, say, Briffa, you get to read it. If you’re wondering why the tone is what it is . . . well, here’s my take.
    .
    I take issue with the following:
    .

    6. Finally, devote followers of this blog seem to be obsessed with an elevated MWP. It is stated so very often that we are “cherry picking” purposely to deflate the MWP. This is simply not the case. In fact, the fatal flaw in this blog and what keeps it from being a useful tool for the palaeoclimatic and other communities is its persistent and totally unnecessary negative tone and attitude, and the assumption that our intention is faulty and biased, which keeps real discourse from taking place.

    .
    No. Many of the regulars here are far less concerned about efforts to “deflate” the MWP as they are about the lack of any credible evidence that the proxies used to reconstruct the MWP are capable of doing so – at least not to within the discrimination claimed (like being able to tell whether current temperatures are “unprecedented”). This is a subtle, but important, difference. I personally really couldn’t care less why you chose Yamal over Polar Urals, and I also would not ascribe any ulterior motive to your choice.
    .
    What I do care about is the cavalier attitude the entire dendro community takes with respect to its work. You realize you have multiple collinear signals (temp and CO2) within each sample and no way to distinguish them, right? You do realize that even split cross validation tests take place during times when these two factors remain collinear, right? You do realize that ring growth is not linear in real life, right? Or that precipitation also affects growth? Or that calibrating based off trees that right now are temperature stressed – but then reconstructing with trees that cannot be shown to also be temperature stressed (and likely aren’t) invalidates the calibration, right? Or that if you have two sets of data – like, say, Polar Urals and Yamal – and they behave differently with respect to each other and local temperature – then that means something is frickin wrong with the theory that allows their use as temperature proxies, right?
    .
    If you wonder why there is so much emotion tied to these discussions . . . well, that’s why. The issue of whether Yamal or Polar Urals (or both . . . or neither) should have been used is just one among many. They are all important issues, since the output of this field has had – and continues to have – a dramatic political influence that will affect every one of us. The answers in both the peer reviewed literature and from dendros themselves (both here and elsewhere) have yet to resolve any of them. The behavior of many individuals in the dendro community (from the outside looking in) invite suspicion. My honest assessment of the community is that (generally speaking) it is deliberately trying to over-reach its ability to draw inferences about past climate by biased sampling and archiving, poor use of statistics, habitual refusal to disclose data and methods, and lack of rigorous procedures for selecting and processing data.
    .
    I imagine this sentiment is shared by many.
    .
    If the dendro community wants the tone to change, then it had better start doing rigorous science.”

    .
    .
    Please take special note of the fact that in reference to Wilson’s “flattening the MWP” comment I specifically answered that was not what I was concerned about. I even put the “deflate” in quotes to emphasize the fact that I have no idea if what they do “deflates” the MWP . . . because I neither know what the “MWP” temperature really was nor do I believe their methods can accurately determine it – yet the community consistently portrays otherwise.

  559. Nathan,
    .
    Meh . . . we cross-posted . . . and though I edited after your post came through, I hadn’t read it through. The dangers of near real-time. 🙂
    .
    The reason I believe it is important is because it is politically important. Scientifically, it’s not the most important thing, but politically it’s critical.
    .
    Scientists don’t make policy. Governments do. And most major governments are democracies/republics. The only way for governments to take the kinds of draconian actions currently being proposed is to ensure the people believe that AGW is both real and comes with catastrophic consequences.
    .
    Hockey sticks are easy to interpret, easy to understand, and quite powerful for that purpose.
    .
    I’m surprised I should have to explain this.

  560. Simon:

    To the extent that the reconstructions are reliable, the temperature change over the whole period is unprecedented, regardless of whether it has been the outcome of natural forcings, anthropogenic forcings or the mischief of the fairies at the bottom of my garden.

    But the point of being unprecedented is to frame CO2 as a clear and present danger to humanity that must be acted on immediately.

    If it’s due to natural causes, well then Rachel’s comments apply: It’s unstoppable and all we can do is ameliorate the effects of climate change, but climate change itself remains (mostly) outside of our control.

  561. “…the student should have access to every data set used to produce the figures in the book, and ideally to more comprehensive data sets that draw the student into further and even original analyses. To this end, I have set as a ground rule that I would not use reproductions of figures from other works, nor would I show any results which the student would not be able to reproduce.”

    bugs,

    Thanks for the link. I looked through it and found the above quote that has some good ground rules about the availability of all data, the reproduction of analyses, and doing one’s own analyses.

    Andrew

  562. Simon:

    To the extent that the reconstructions are reliable, the temperature change over the whole period is unprecedented, regardless of whether it has been the outcome of natural forcings, anthropogenic forcings or the mischief of the fairies at the bottom of my garden.

    .
    Really? So you believe that the methods used to reconstruct the past 2 millenia have the discrimination to show that the 1850-2010 temperature change is unprecedented?
    .
    And given that this temperature change is a 150-year episode (kinda-sorta), do you really think that a sampling period of only about 13.5 times that period is sufficient to exclude the possibility that it is within the range of natural variability?
    .
    If so, then that is where we disagree sharply.

  563. Nathan,
    .
    “It would seem Raven would rather ignore all the knowledge we have because that knowledge isn’t perfect.”
    .
    Please read what I write instead of constructing strawman. All knowledge is incomplete and it is possible to make decisions in the face of uncertainty. However, the level of uncertainty affects the choices that we make. If there is a lot of uncertainty then some decisions cannot be justified because the cost is too large to offset hypothetical benefits.
    .
    For example, you do not know if your house will burn down but it is certainly a plausible risk. You would also think nothing of spending a $1000/year to protect yourself against that risk. However, if the insurance was going to cost your $100,000/year there is a good chance that you would live with the risk and hope for the best because the cost of insurance is too high.
    .
    That is the situation we are in with CO2 because we do not have the technology that will allow us to maintain the societies we have now without emitting a lot of CO2. If we knew for certain that catastrophic climate change was coming in 50 years then we might be able to justify the huge sacrifices that would be required to move to low emissions technologies. But the reality is we don’t know. Catastrophe might take 100 or even 500 years to arrive or maybe it will never come. Given that unknown we cannot justify a made rush to meet draconican CO2 emissions target in the short term. That is also why the existance of the MWP affects our decision because it would demonstrate what most people already know: that the alarmists are exagerrating the knowledge we have about climate.

  564. RyanO

    “Scientifically, it’s not the most important thing, but politically it’s critical.”

    I don’t even think it’s politically that important. The claims that there was a need to take action started well before the Hockey Stick. Remember Kyoto?

    Case for action in 1987:
    http://muse.lib.nchu.edu.tw:97…..8123a0.pdf

    It is important to make the distinction between what is politically important and what is scientifically important.
    The case to act on AGW is based in science, radiative science. The political case to act on AGW is… well who cares really? Because the science case is what is important, no?
    In fact by focussing on what is politically important and attempting to cast doubt (or casting doubt as the case may be) on solely that aspect you reveal that your intention is to influence the political debate. Why would you focus on the political debate when the case to act is in the scientific debate?

    And can I say that on CA, when Rob Wilson posted to explain his actions the response was pretty lame. If you want to genuinely change the way these dendro’s do their work do you shout at them “You’re doing it wrong, your’e not doing science” whenever they drop by? It’s just a bad tactic. Especially when you are saying that not only do you not like their results, but they have to do it a way which meets your approval. Why not do it yourself, enough with the moaning. Why not get involved… Actually you are, with that Antarctic thing. Sorry, Kudos to you.

    “The only way for governments to take the kinds of draconian actions currently being proposed is to ensure the people believe that AGW is both real and comes with catastrophic consequences.”
    yeah, see… This is just not anything to do with the science is it? If you don’t want to do anything about AGW, then make a case for it. Sitting around saying the proxy reconstructions aren’t as good as you’d like is not a particularly good argument.

  565. Raven

    “That is the situation we are in with CO2 because we do not have the technology that will allow us to maintain the societies we have now without emitting a lot of CO2. If we knew for certain that catastrophic climate change was coming in 50 years then we might be able to justify the huge sacrifices that would be required to move to low emissions technologies. But the reality is we don’t know. Catastrophe might take 100 or even 500 years to arrive or maybe it will never come. Given that unknown we cannot justify a made rush to meet draconican CO2 emissions target in the short term. That is also why the existance of the MWP affects our decision because it would demonstrate what most people already know: that the alarmists are exagerrating the knowledge we have about climate.”

    You just made all that up. Have you done some economic modelling? Or are you just personally alarmed becuase you don’t like paying taxes?

    All you are doing is replacing the current best guess, with “we don’t know”. Frankly that’s a little less than convincing. Heck according to you it’s just as likely to be worse. To really combat AGW you need to have a better scientific theory, not better political tactics.

    “Given that unknown we cannot justify a made rush to meet draconican CO2 emissions target in the short term. ”
    What mad rush? The targets are actually pretty low… Currently no target at all. Australia is looking at a 5% reduction from 2000 Levels…. Ummm I don’t think you get much lower as a target than that?

    “That is also why the existance of the MWP affects our decision because it would demonstrate what most people already know: that the alarmists are exagerrating the knowledge we have about climate.”

    So go and find it. Stop whining that people won’t find it for you. The people doing reconstructions don’t seem to be finding it. Go and collect your own data and find it.

  566. Nathan,
    .
    Yeah. And Kyoto really went places, didn’t it? And why didn’t it go places, Nathan?
    .
    You completely, utterly misunderstand my point – and I think it’s deliberate. I have no desire to play politics. My frustration is that others are playing politics by doing shitty science, and the shitty science has had a definite effect on public opinion – an effect that (fortunately) seems to be waning in recent years.
    .
    You know darn well that no one who has been arguing with you is advocating replacing science with politics.
    .
    And if the dendros methods are crappy (and they are), then the appropriate response is to make that fact known to them and others. Hence the topics at CA. And yes, Nathan, if you genuinely want to make someone change how they do things, you either:
    .
    1. Show them how to do it right; or,
    2. Tell them what they need to do to have a prayer of doing it right
    .
    Since it cannot be shown that a “right” (and by that I mean one which is supported by a physically valid theory) way exists, then the latter is most definitely appropriate. You can call it lame all you want, but until they can quantify the list below (along with other things) it’s not rigorous science. It’s wishful thinking. Or sorcery. Or something.
    .

    What I do care about is the cavalier attitude the entire dendro community takes with respect to its work. You realize you have multiple collinear signals (temp and CO2) within each sample and no way to distinguish them, right? You do realize that even split cross validation tests take place during times when these two factors remain collinear, right? You do realize that ring growth is not linear in real life, right? Or that precipitation also affects growth? Or that calibrating based off trees that right now are temperature stressed – but then reconstructing with trees that cannot be shown to also be temperature stressed (and likely aren’t) invalidates the calibration, right? Or that if you have two sets of data – like, say, Polar Urals and Yamal – and they behave differently with respect to each other and local temperature – then that means something is frickin wrong with the theory that allows their use as temperature proxies, right?

    .
    Your continued refrain is, “if you don’t like the reconstructions, do one of your own.” Your assumption is that it is possible to do so. You seem to be deliberately avoiding the statements by myself and others that the reason we don’t do them is because we don’t believe the results have any physical meaning.

  567. Simon Evans, comment 22321
    “The GCMs actually reproduce most of the 1910-1940 period quite readily through natural forcings only, and some runs will, of course, show temperature change in excess of the record for that period. What they will not do is reproduce the whole of the 20th century without anthropogenic forcings”.

    You’re referring of course to the 2 graphs in IPPC-4 WG1, Chapter 3, 2007. Labeled “Comparison between computer simulation and observation and observation of the global surface temperatures (deg C), from observations (thick-black) and computer simulations (dark grey) using (a) both human and natural factors and (b) natural factors only.

    May I quote from Peter Taylor, in his book, “Chill, A reassessment of global warming theory”, Clairview, 2009, page 27.
    “However that edifice is now beginning to crumble. Not only is it clear that natural factors were not known enough to be modeled, as the US National Academy of Sciences suspected, but recent work has shown that the models falsely replicated key elements of the past pattern. In particular the ‘global dimming’ period of falling temperatures between 1945 and 1978 was assumed to be caused by sulphur particulars from fossil fuel emissions. The models incorporated assumptions about the power of sulphate aerosol to create the dimming, resulting in 3 decades of cooling despite the increases in carbon dioxide. The models also built in erroneous assumptions about upper ocean heat storage derived from a monitoring system now known to be flawed. Models replicated the ocean warming that had been reported, but the later work showed the reality had been 200% less”

    And on pages 54/55 in relation to the sulphate dimming hypothesis; –
    “Scientific opinion has, however, now shifted in the last few years. It is now clear that sulphate pollution from either industrial emissions or other sources such as volcanoes could not have been responsible for the dimming because it is too localised. The most recent reviews of satellite data show that changes in natural aerosols and cloud patterns are implicated and that attributing the source of ‘global dimming’ to industrial aerosols was led by an artifact of measurement protocols that were biased to land and certain polluted and certain pollutes regions of the northern hemisphere.”

  568. RyanO

    I suspect you are playing politics too.
    “You completely, utterly misunderstand my point – and I think it’s deliberate. I have no desire to play politics. My frustration is that others are playing politics by doing shitty science, and the shitty science has had a definite effect on public opinion – an effect that (fortunately) seems to be waning in recent years.”

    This makes no sense though. As I had hoped to demonstrate to you that the Science still demands we take action. You can argue all you like about what is the most effective political tool, it matters nothing. Because the science, the radiative physics, demands that we take action. UNLESS you simultaneously have a problem with that aspect of AGW – do you?

    “Yeah. And Kyoto really went places, didn’t it? And why didn’t it go places, Nathan?”
    Well, it was a starting point. I don’t support your argument that the current political will to do something (but almost nothing) about AGW is derived solely (or even in a large-ish part) from the Hockey Stick. The most common graph we see showing temps is the one that just shows the actual temperature record (from either 1979 or around 1850)

    “Your continued refrain is, “if you don’t like the reconstructions, do one of your own.” Your assumption is that it is possible to do so. You seem to be deliberately avoiding the statements by myself and others that the reason we don’t do them is because we don’t believe the results have any physical meaning.”

    Ok, so write a paper demonstrating they have no physical meaning. If you can’t write why they have no physical meaning – or why there can be no confidence they have found a physical meaning – then where is your argument?

    “2. Tell them what they need to do to have a prayer of doing it right”
    So, why don’t you publish that work? Why don’t you make AT LEAST a summary of what needs to be done. And if you want people to take on board your criticisms, do yo ureally think you’ll do that by telling their current results are nonsense and that they are “deliberately trying to over-reach [their] ability to draw inferences about past climate by biased sampling and archiving, poor use of statistics, habitual refusal to disclose data and methods, and lack of rigorous procedures for selecting and processing data.”

    This is just nasty.

  569. Further to the above. This modeling also apparently only uses TSI for the sun’s influence. Recent work has lead to a large change in the variation of TSI over the 20th century. See the 3 graphs in the link, Lean 2000, Wang 2005 and Leif 2007. Another flaw.

    http://www.leif.org/research/TSI-LEIF.pdf

    Apparently even Lean doesn’t believe his graph anymore.

  570. Geoff,
    Simon’s defense here is weak in other ways too. I would discuss it with him in a quiet thread somewhere, but not where Nathan is present to noisy the signal. It’s nothing I haven’t mentioned already, but I suspect Simon doesn’t read everything I write. (Which is fine. No snark.)

  572. RyanO

    It’s also strange that you simultaneously claim:

    “the reason we don’t do them is because we don’t believe the results have any physical meaning”

    and

    “2. Tell them what they need to do to have a prayer of doing it right”

    So you don’t think they have any physical meaning, but somehow have the understanding of how to do it right?

    If you know how to have a prayer of doing it right, why can’t you do it right and show them the result.

  573. His work doesn’t “prove” there is anything “fishy” going on. You do recognize this is a publication in Science, right? You have to read between the lines, right? Why did Broecker write the paper? What did he say? What was the reply? Remember: these are academics writing in Science, not me and you slugging it out in the blogospehre. If you can’t appreciate nuance, the message will be lost on you.

  574. Did Broecker (2001) find Mann’s (1998) hockey stick credible? Contrast for me his response versus, say, McIntyre’s.

  575. Bender

    I posted what I thought above:

    Well Broeckers paper was very nice… He doesn’t really get very specific. Was there something in it you wanted me to see specifically?
    “The case for a global Medieval Warm Period
    admittedly remains inconclusive. But keeping
    in mind that most proxies do not have adequate
    sensitivity, it is interesting that those capable
    of resolving temperature changes of less
    than 1 “C yield results consistent with a global
    Medieval Warm Period.”
    That’s the opening of his conclusion.

    “I consider this evidence sufficiently
    convincing to merit an intensification of
    studies aimed at elucidating Holocene climate
    fluctuations, upon which the warming
    due to greenhouse gases is superimposed.”
    That’s the end. So he says we need further work (this is around 2001)

    Did he do anymore work?

    Then I looked for his more recent work and mostly he seems to be supporting action against climate change. Or I should say Global Warming as apparently he coined the term.

    I didn’t find the reply. Do you have a link?

  576. lucia, discussing MWP and HS is way OT for this thread. You ok with that? You want we pick a different thread?

  579. Bender

    People can doubt Mann’s Hockey Stick, it’s fine. And Broecker here outlined that he didn’t think it was accurate. And then he went on to describe what he thought was the real deal. Good on him, but… Has he continued along that path? Has he done more recent studies that also confirm the MWP? Or that there is the 1500 year THC cycle?

    I don’t see what’s so significant.

    Did he get beyong this point:

    “I consider this evidence sufficiently
    convincing to merit an intensification of
    studies aimed at elucidating Holocene climate
    fluctuations, upon which the warming
    due to greenhouse gases is superimposed.”

  580. Nathan,
    .
    “You just made all that up. Have you done some economic modelling? Or are you just personally alarmed becuase you don’t like paying taxes?””
    .
    First, you need to learn something about economics. If non-emitting energy sources require tax dollars then they are unsustainable. Non-emitting energy sources need to be pay for themselves.
    .
    The biggest problem is the replacement technologies don’t exist. Wind and solar are unreliable and incredibly expensive and the physical size of the infrastructure means that savings from ‘economies of scale’ are limited. Nuclear would help but nobody wants to let the plants get built. Most of the good hydro resources are already exploited. We need coal or natural gas but gas is also an expensive way to produce electricty.
    .
    The AGW alarmists are living in a fantasyland if they believe that we can replace fossil fuels over the next 20-50 years. 100 years might be possible depending on what technological advancements occur. If the science really says we need have zero CO2 emissions by 2050 to avoid ‘dangerous climate change’ then we are screwed and mitigation is a waste of time.

  581. Nathan,
    .
    Science demands that we take action? What?
    .
    What action does science “demand” we take?
    .
    The issue is not whether GHGs increase SAT. I accept that they do. The issue is magnitude. The science does not demand that we do anything to prevent certain apocalypse because the science does not tell us that apocalypse is certain.
    .
    And your constant refrain to publish is getting really tiresome. If I am right, I am right – regardless of whether I publish. If I would rather post on blogs than publish in journals, so what? How does that make what I say any less right (or wrong)? You seem to have a rather exaggerated idea of the impact simply “publishing” has. A paper is static, makes no noise, and can easily be ignored. What keeps papers alive is discussion – be it at blogs, conferences, email, lectures, or any other forum. It is the discussions that are the most important. That is where all the convincing is done. Papers are just a way to formalize it.
    .
    For the last . . . I really don’t care whether you think it’s nasty because I see no reason to sugar-coat the truth. They do overreach in their ability to draw inferences about past climate. They do intentionally and unintentionally bias their samples, use shoddy sampling techniques, and deliberately fail to archive data and metadata. They do repeatedly demonstrate poor or incorrect use of statistics. They do refuse to divulge methods, codes, and data. They do use constantly changing, ad hoc methods to select samples and process data, including making decisions ex post that must be made ex ante. Nasty? Nasty science, maybe.
    .
    You also seem to forget that my post didn’t come out of nowhere; it was Wilson who made a slightly petulant post about how CA and its readers could do a better job with a somewhat accusatory assumption as to why many of us get all a-blather about tree rings.

  582. Nathan,
    .

    So you don’t think they have any physical meaning, but somehow have the understanding of how to do it right?

    If you know how to have a prayer of doing it right, why can’t you do it right and show them the result.

    .
    Dude, seriously. The physical theory bestows the physical meaning. The math is the how. Even if they do all the math right (and stop doing hoaky things like RCS and CPS), that still doesn’t mean that their results have physical meaning.

  583. RyanO

    “Science demands that we take action? What?”
    Ok, make that the demand for action is rooted in science.

    “The issue is not whether GHGs increase SAT. I accept that they do. The issue is magnitude. The science does not demand that we do anything to prevent certain apocalypse because the science does not tell us that apocalypse is certain.”

    So your whole argument is that because we can’t be certain what the magnitude is, we do nothing. Ok fine.

    the refrain publish will go on. Because publishing is important. Can you really expect dendro’s to wander over the CA and find out how to do it right? Will they be using references to CA in their papers? Will they be using CA as a teaching aid in lectures?

    You also didn’t adress this contradiction:

    “the reason we don’t do them is because we don’t believe the results have any physical meaning”

    and

    “2. Tell them what they need to do to have a prayer of doing it right”

    So you don’t think they have any physical meaning, but somehow have the understanding of how to do it right?

    If you know how to have a prayer of doing it right, why can’t you do it right and show them the result.

  584. And if you want to have the last word, go nuts . . . I will post again if things get back to technical stuff.

  585. “You also seem to forget that my post didn’t come out of nowhere; it was Wilson who made a slightly petulant post about how CA and its readers could do a better job with a somewhat accusatory assumption as to why many of us get all a-blather about tree rings.”

    I think I was right about why you all get ablather about tree rings, it has to with $$$$. You want to affect a political change – it really has nothing to do with science, you just want to avoid mitigation.

  586. Nathan, if your mind is made up, then what exactly are you doing here trying to convert people a who appear to be a whole lot smarter than you? If you believe in action-damn-the-facts, then you hzve no place here. If you have questions, ask them.
    .
    Re: Broecker (2001). I am not referencing a person. I am referencing a paper. What would motivate someone to write the words that were written? He clearly found the Mann reconstruction INcredible, for the absence of a MWP. McIntyre too. Broecker’s response was to try to explain how the MWP had to be so significant and large-scale (if not global) that it could not be disappeared except through voodoo. McIntyre found the source of the voodoo. And he did it again, And again. And again. And again. (1) Where are all these busted hockey sticks coming from and (2) why are they so easily broken by someone with knowledge, insight, and analytical skill?

  587. Raven

    The Stern Report, the Garnaut Report (here is Australia) both say that mitigation now is the economically correct thing to do.

    “The AGW alarmists are living in a fantasyland if they believe that we can replace fossil fuels over the next 20-50 years. 100 years might be possible depending on what technological advancements occur. If the science really says we need have zero CO2 emissions by 2050 to avoid ‘dangerous climate change’ then we are screwed”

    well good, give up then. Don’t even try… I mean the CSIRO here in Australia certainly don;t have any plans, they’re certainly not trying to make polymer solar films. As cheap as money to print and able to stick on pretty much anything… Nope that won’t happen. Can’t build windfarms with natural gas back-up. Nope that’s just a pie in the sky.

    I am glad we had this discussion.

    I wonder why you all at CA think you are capable of ‘auditing’ the work of proxy reconstructions fairly when it’s pretty clear you have a reconstruction you want to see or expect to see. This is a clear bias. Your aim does seem to be to keep the MWP warm to affect a political change.

  588. “I think I was right about why you all get ablather about tree rings, it has to with $$$$. You want to affect a political change – it really has nothing to do with science, you just want to avoid mitigation.”
    .
    Like we didn’t know your opinion the moment you arrived and started shooting off your mouth. But you are entitled to your opinion. It makes you unworthy of further response, however.
    Good night.

  589. Nathan,
    .
    “The Stern Report, the Garnaut Report (here is Australia) both say that mitigation now is the economically correct thing to do.” .
    .
    Of course they did since these reports were developed to justify policy choices that were already decided in advance. Be honest now: do you really believe that Garnaut would have said anything elese given the fact that Rudd had already committed carbon mitigation?
    .
    These reports also rely on the assumption that we can predict what the future “costs of adaptation” will be. We have no idea what those costs might be because we have no idea what will happen to the climate.
    .
    Here is a much better report on the technologies available to us:
    http://www.withouthotair.com/
    The author has managed to cobble together a set of plausible assumptions regarding technology that could, in theory, provide GHG free energy but the cost of building and deploying the infrastructure required would be astronomical.

  590. “Your aim does seem to be to keep the MWP warm to affect a political change.”
    .
    The aim is to get the science right and let the chips fall where they may. The contamination of science by political motivations – by religious zealots such as yourself who don’t give a hoot about the scientific and economic facts – is requiring extraordinary efforts to keep the science on track.

  591. Nathan,
    .
    “I wonder why you all at CA think you are capable of ‘auditing’ the work of proxy reconstructions fairly when it’s pretty clear you have a reconstruction you want to see or expect to see.”
    .
    When Mann 2008 came out I was prepared to accept his results of a MWP that was no warmer than the 1980s. I even posted a comment on CA that when something like: “I expect Mann learned something and the algorithms will likely hold up to scrutiny”.
    .
    I was shocked to find out how wrong I was. The problems with Mann 2008 are symptomatic of a deep rot in climate science that taints everyone involved even if they are not to blame.

  592. Raven,

    Regarding the notion that if we cannot replace emissions producing energy sources within 50 years then we might as well do nothing, this is a pretty silly argument and one of the reasons why sometimes I find it very difficult to take the position of those who doubt AGW seriously.

    If we start with the assumption that two degrees of warming will have bad effects and that three degrees of warming will have worse effects and so on and so on, then even if we cannot avoid two degrees, avoiding three degrees is obviously desirable. Thus, even if we cannot succeed in stopping emissions within 50 years, stopping them within 100 years would be a very good thing to do.

    There sometimes seems to be a progression in the arguments used by some people along these lines:

    1.) There is no warming.
    2.) There may be warming, but we’re not causing it.
    3.) We might be causing it, but it’s not a problem
    4.) If it is a problem, we can’t do anything about it.

    While you have not gone down this path, you certainly seem to have jumped to position 4. Am I misrepresenting you here?

  593. Bender

    I don’t care if you’re smarter than me. Big deal. Nor how you determine that.

    “If you believe in action-damn-the-facts, then you hzve no place here.”
    What? You don’t have any facts. Your case revolves around ‘we don’t know enough’.

    “Re: Broecker (2001). I am not referencing a person. I am referencing a paper. What would motivate someone to write the words that were written? ”
    So you went from referencing the paper, to the person. You are asking about his motivation. You are referencing a person, this person:

    http://en.wikipedia.org/wiki/Wallace_Smith_Broecker

    “In 1975, Dr. Broecker inadvertently coined the phrase global warming when he published a paper titled: “Climate Change: Are we on the Brink of a Pronounced Global Warming?”[7] He has recently co-written an account of climate science with the science journalist, Robert Kunzig. This includes a discussion of the work of Broecker’s Columbia colleague Klaus Lackner in capturing CO2 from the atmosphere – which Broecker believes must play a vital role in reducing emissions and countering global warming. Broecker has been described in the New York Times as a geoengineering pioneer.[8]”

    Quite clearly he is very upset about the Hockey Stick and thinks we should take no action. Of course when he wrote that Science article he was amazed about the lack of MWP and new that the Voodoo being used was political mischief. 🙂 YEAH!

    “He clearly found the Mann reconstruction INcredible, for the absence of a MWP. McIntyre too. Broecker’s response was to try to explain how the MWP had to be so significant and large-scale (if not global) that it could not be disappeared except through voodoo. ”

    That’s a very interesting interpretation. I don’t remember seeing the word Voodoo anywhere. And he’s such a vocal voice against those nasty Team members…

    What exact part of the paper leads you to think there is voodoo (or naughty tricks)?

    “McIntyre found the source of the voodoo. And he did it again, And again. And again. And again. (1) Where are all these busted hockey sticks coming from and (2) why are they so easily broken by someone with knowledge, insight, and analytical skill?”

    Well that doesn’t make much sense. If it was so easily busted why do they keep appearing? Because they weren’t actually busted? Becuase the authors aren’t aware of the bustedness becuase McIntyre doesn’t publish why each one is busted…

  594. A religious zealot that can’t think straight. What a surprise. Bye-bye already. “Thanks for playing”, as our friends at realclimate like to say.

  595. David,
    .
    “Regarding the notion that if we cannot replace emissions producing energy sources within 50 years then we might as well do nothing, this is a pretty silly argument”
    .
    I can support a mitigation strategy as long as it does not include any arbitrary (and unobtainable) targets/timetables, carbon trading or bans on nuclear. It could include a modest carbon tax, subsidies for promising technologies and regulations designed to improve efficiency.
    .
    I end up opposing mitigation strategies because anyone who supports them invariably insists on binding targets and carbon trading.

  596. Raven,

    That does not really seem to address the issue that I had with your post, which was that your argument was silly on its face. The argument that I referred to was:

    “If the science really says we need have zero CO2 emissions by 2050 to avoid ‘dangerous climate change’ then we are screwed and mitigation is a waste of time.”

    Re bans on nuclear, I would not ban nuclear. It is going to be part of the solution.

    However, if the science said that x needed to be done in order to avoid y, and x was a limit on carbon emissions, what would be ‘arbitrary’ about that?

    On carbon trading versus a carbon tax, I do not really have a position. Given that a carbon tax is not on the table in Australia in the policy positions of any of the major parties, the issue is not on my radar.

  597. Simon Evans,

    Would 20 degrees over 50 years qualify as unprecendented??

    http://www.uswaternews.com/archives/arcglobal/8icecor11.html

    http://www.sciencemag.org/cgi/content/abstract/sci;282/5386/92?maxtoshow=&HITS=10&hits=10&RESULTFORMAT=&fulltext=50+years&searchid=1&FIRSTINDEX=0&issue=5386&resourcetype=HWCIT

    Onward, our hostess likes to work with the last 10 years. One of the important reasons for this is that there has NOT been a major low latitude volcanic eruption that went stratospheric since the early 90’s. There have been few large eruptions.

    This means that neither natural nor human produced sulfur compunds are hanging around in large amounts to suppress AGW. Funny thing is, temps have been flat for this time period.

    Clear skies to get that solar insolation into the oceans and ground. Lots of built up energy in the oceans we are told. Co2 continuing to rise. We are assured that there is warming in the pipeline. Why hasn’t it exhibited?? Where is it?? What is the mechanism that is suppressing it???

    This would tend to make Dr. Spencer sound quite credible at this point. That is, the most likely mechanism for the temp change is cloud cover. As the temps are flat, it would tend to make the idea of positive cloud feedbacks false.

    I am personally getting tired of scientists claiming extraordinary hidden mechanisms to cover holes in their theories. Unknown mechanisms are not testable among other things.

  598. kuhnkat

    The solar cycle may well be an explanation for some of the hidden warming over that period. Solar cycle 23 peaked in 2000. It reached a minimum in 2007 and has stayed there ever since. From my limited reading on Wikipedia ;), it would seem that the peak to trough difference in forcing is probably sufficient to hide 10 years of global warming.

    While solar cycles average 11 years, or 5.5 years from peak to trough, this one has gone longer, with the trough extended. When you add a la nino to that in 2008, my suspicion is that we basically have our explanation.

  599. David,
    .
    “However, if the science said that x needed to be done in order to avoid y, and x was a limit on carbon emissions, what would be ‘arbitrary’ about that?”
    .
    1) Because the science cannot provide that level of certainty so any number picked is arbitrary even if it has a psueudo-scientific rational.
    .
    2) Because the cost effective emission free technology does not exist so it will likely be impossible to meet any target set in the manner you describe.
    .
    Carbon trading is scam and will never work because everyone involved has an incentive to cheat and to allow others to cheat. Anyone who promotes carbon trading as a solution is either naive or simply a carpet bagger looking to make easy money at the expense of others.

  600. The aim is to get the science right and let the chips fall where they may. The contamination of science by political motivations – by religious zealots such as yourself who don’t give a hoot about the scientific and economic facts – is requiring extraordinary efforts to keep the science on track.

    Broken my irony meter again.

  601. David Gould,

    talk to Leif Svalgaard, he keeps assuring us over at WUWT that it can not be counted for the difference up or down. I would point out that the temps have been going nowhere since the MAX, NOT the min and that the area under the curve is huge compared to the tiny changes on the top that get graphed as if they are large.

    That is, the TSI ranges from about 1316-1408 over a yearly period. The large variation is NOT due to the solar cycle. It is due to orbital eccentricity.

    Here is the data for SORCE:

    http://lasp.colorado.edu/sorce/tsi_data/daily/sorce_tsi_L3_c24h_m29_v09_20030225_20091015.txt

    Column 5 shows TSI at 1AU, the number you watch if you are studying the sun.

    Column 10 is the TSI at Earth Distance.

    (SORCE does not cover a full cycle)

    The sun varied about 2 w/m2 over cycle 23:

    http://www.pmodwrc.ch/pmod.php?topic=tsi/virgo/proj_space_virgo

    Darn few people who can actually calculate this stuff think that this small variation accounts for the climate. That is why Svensmark’s finding is important. It takes a little more away from the forcings that get swept into the GHG bin.

    It is also why the alarmists resist Dr. Spencer’s work so much. His negative feedback findings with clouds conceivably rearranges a LOT of forcing. It could also salvage the idea that GHG’s heat the earth without there being a warming problem. Negative cloud forcings balance the GHG positive forcing.

    http://www.drroyspencer.com/

    By the way, less TSI wouldn’t hide anything. If you reduce amplifier power, are you HIDING an increase in volume??? No, the energy is not entering the system to provide signal amplification. The best that can said, in that situation, is that the ability to amplify the signal when the power is returned is still there!!

    Finally, do not confuse El Nino’s, La Nina’s, ENSO, AMO… with DRIVERS like the sun. For instance, the wind drives a windmill producing power, but, the air molecules do not produce the momentum they carry. That is forced into them from an outside energy source whether it is gravity or thermal energy.

    Similarly, those ocean cycles do not produce the energy, they simply show where it is, or is not, moving through the system.

  603. It is funny Bugs.

    Bender seemed to have a strange like of the Broecker paper, said it somehow proved what he thought, but never actually quoted which part.
    “Re: Broecker (2001). I am not referencing a person. I am referencing a paper. What would motivate someone to write the words that were written?”

    Have you looked at what Broecker wrote?

    Better still Look at Wallace S. Broecker’s wiki entry, it’s hilarious that he’s Bender’s darling defeater of hockey sticks…

  605. Pingback: Don’t Hold Your Breath. If the AGW Evidence is So Conclusive, AGW Proponents Explain This: « Thoughtful Analysis

Comments are closed.