The Blackboard

Every one a winner. Don’t fit, change it, does fit, claim you knew it would and now you can”project” to the year 1,000,001 and be certain of 100% right.

[...] stimulating climategate, Follow the political money trail, the times they are a changing, window shopping for models, hurricanes not really in a [...]

I’m really really good at modeling the past. And I can also model the future , but it takes time and the results might not be available until the future is the past and I have had time to tweak them to get a good fit. I can then use the same model to predict the future again, but a little more tweaking might be required. etc. etc.

I bet 100 quatloos on the December UAH anomaly to be . . . 0.280 C. I win!!! Please update my account!

Thanks!

In my work I have had done some modeling based on water vapor radiometer data for telecom work. In order to test my models, I had used statistics from past readings, made a prediction and tested the validity of those predictions by testing it on some of the data from the same WVR that wasn’t used for statistics. Why aren’t we doing that for climate models? That is, why aren’t we using the data from say anything before 1995 to make our models and then see if the models match what we observe from 1995 to the present? What is so difficult with that?

This is an honest question, and I would love to get an answer for this.

Shervin–
That method of testing doesn’t quite work for climate models because they aren’t curve fits in any traditional sense. (Some of the parameterizations use curvefits of various sorts to certain data, but they aren’t curve fits to ‘earth’s surface temperature’. They are curve fits of the sorts commonly used by engineers to create simplified computational fluid dynamics or heat transfer codes.)

In principle, modelers could run their AOGCMs through 1990, test, pitch the AOGCMS that did badly and start over. Then run the successful ones through 2000, see how those work, and use the ones that passed both tests to project. The problem is that the models take so long to run they really can’t do that.

But the short answer to your question is: They can’t do what you do because the method of projecting is fundamentally different. The method of developing projections is not inherently a statistical fit and it’s also much more computationally intensive.

Are you talking about Tamino’s post where he compares observations between 1900 and 2010 with models, notes good agreement, then notes some models are much cooler prior to 1960 (which means an overall higher warmer trends), discards those results and shows that observations between 1900 and 2010 fit much better.

If so I used to think Tamino was over reacting with his attacks on you, and now I don’t.

You write: “Specifically, Tamino threw out the a model with a high trend after 2000 from the suite of models used by the authors of the IPCC AR4 when making projections in the .” I suspect that a link is missing here.

Lucia,

Did you see now NASA and NOAA have now joined CRU in the newest “Climategate”?
http://www.spaceref.com/news/viewpr.html?pid=30000
http://icecap.us/images/upload.....tegate.pdf

Michael–
I don’t know if Tamino is over-reacting to me or not. That’s not really my concern. But he’s being silly.

With respect to what you think Tamino showed, could you quantify “good agreement”? I don’t consider the agreement he showed good. I think he is showing the same graphs that don’t do a good job of really showing whether or not models and observations agree very well. (Or at least, can’t show better than qualitative agreement. )

If you compare the trend computed since 1950 without throwing out data, the model means trend is higher than the observations and the result is statistically significant. I showed that last week. The differences can be about 10%-50% of the average predicted warming (or more depending on how you look at it.)

The models don’t appear to be doing much better than linear extrapolation. So, given their complexity, how is that good agreement?

Now, since some models overpredict warming trends by a factor of more than 2 and some under predict warming by a factor of more than two, and the temperature have been baselined so that the are mathematically forced to fit between 1980-1999, if you plot temperatures vs. time and compare observations and simulations, some do fall below the models that are too high and below those that are too low. No one has ever claimed otherwise.

But with a spread in temperature like that and a mathematical technique that forces a fit for part of the time period, how could the observations not fall inside the range? By what definition is merely falling somewhere inside the range evidence the agreement between models and observations is “good”?

Given that huge spread, the IPCC suggested we could be guided by the multi-model mean and that it’s the multi-model mean whose fit should be good. IF that’s their advice, it seems to me that when testing their guidances… well… test what they actually suggested we should expect. That is: Test whether the multi-model mean is on track.

It’s not. The multi-model mean is high. The difference is statistically significant. So, presumably, at least some of the models are too high (for whatever reason. Bad physics, bad forcings? I don’t know.)

Now, it goes without saying that with a spread like that, if one throws out model runs after observations are collected, you can get better agreement. If the observations had come in low, we could have thrown out runs from Echo G which has a low trend. If we compare the trend since 1950, and use ratio, it’s as far off from the model mean as CCMA which Tamino toss out.

If one is going to toss models, why not toss that one out? In reality, the IPCC didn’t toss any out.

I don’t know why Tamino is now throwing out models with high trends the mean trend is lower and fits the new data better. But it’s hardly surprising that if the new data are lower than the mean that includes all the models and you toss out the projections based on models with the higher trends, the multi-model mean will fit the data better than you had anticipatd before you knew what the observations were.

Lucia,

I think Tamino is saying that it is the models that we know to be poor because of their performance in reproducing observations prior to 1960 that are showing the poorest agreement with post 1960 observations. If we remove these models, then we get better results – obviously. But the justification for removing these models is not the poor agreement that they have shown with observations recently (at least not that alone).

Lucia, I really enjoy your blog, and I know that you and Tamino are having a little vendetta going. But here I think you are being unfair to him. I did not read his post as suggesting that the agreement is good after throwing out some models (which would of course be cherry picking), but rather that it is a certain class of models that leads appears to be a consistent cause of the mismatches we are observing.

But with a spread in temperature like that and a mathematical technique that forces a fit for part of the time period, how could the observations not fall inside the range? By what definition is merely falling somewhere inside the range evidence the agreement between models and observations is “good”?

I thought a model collection had to be significantly wrong for 30 years after the latest forced fit before it was actually wrong. As long as the collection is refreshed more often than that, it cannot ever be wrong.

If a duck requires major adjustments to exist, is it still a duck?

David Gould–
But he didn’t remove the models that had the poorest results in the past. He removed the models that were poor and on the high side and he is developing his criteria knowing the outcome of data the IPCC hoped to predict.

In any case, the information these models have poor hindcasts was readily available in 2006-2007 and was not used by the IPCC authors to create their projections. So ending with statements like “putting the lie to claims that recent observations somehow “falsify” IPCC model results.”

Let’s set aside who or what precise claim Tamino thinks he is rebutting and look at the logic of what Tamino seems to be claiming.

Tamino appears now saying the IPPC model results are not “false” because their lack of fidelity with the future arises from inclusion of a particular model which appears to be false. Moreover, that appeared false even before the IPCC authors decided to use it.

Now that we have more data, we discover that it continues to appear false. Now, if, he treats this models as false, rather than true (as the IPCC did) then, he gets a “projection” that looks like it’s not false.

But then he suggest that the fact that he can get good agreement by treating the results of an IPCC model that looks false to him as false… puts “the lie to claims that recent observations somehow “falsify” IPCC model results.”?!

Sheesh!

Lucia,

Which models show poorer results in the past than the ones Tamino looked at? (not snark – I do not know).

And I agree that falsifying the IPCC projections involves falsifying the IPCC projections, not a subset.

Paz–
If he did not end his post with ““putting the lie to claims that recent observations somehow “falsify” IPCC model results.” I would read his post somewhat similarly to the way you do.

However, it’s ridiculous to write a post making a case that one or some of the models are false, and showing that if one eliminates the false models from the groups, and then end it with the claim “putting the lie to claims that recent observations somehow “falsify” IPCC model results.”

Regardless of how those above wish to justify Tamino’s actions, the byproducts of his tamperings are not the results projected by the IPCC.

David Gould (Comment#30214)
I was going by “since the 50s” and by ratio with the average model. (I think rations make more sense than absolute.) By this measure, echo G is worse than CCMA.

science without statistics
bears no fruit
statistics without science
has no root

PS: Better make a new tag called “Statistricks”

My understanding of the issue is that the IPCC has included a wide range of models. The rational was apprently that each group of models did something well that the others didn’t. This wide range of models gave the multi-model mean wide error bars that kept it from being too far from reality. This also allowed them to have a mean projected temp of about 2-3c/C.

The current situation seems that Tamino is finally accepting that the high temp models are wrong and should be tossed. He then does this, yet still claims the IPCC projections that depend on including these hot models are good.

Is he claiming that the projected trend is the same, now that he has tossed the hot runs, compared to the IPCC’s?? That would obviously be a BONER!!!

lucia,

However, tamino is looking at a longer model output/observations comparison than that. So, over the period that he looks at – 1880 onwards – are there any worse models than CCMA? In other words, is tamino removing the worst models under his metric, or not?

Shervin (Comment#30188) January 14th, 2010 at 3:13 pm

In my work I have had done some modeling based on water vapor radiometer data for telecom work. In order to test my models, I had used statistics from past readings, made a prediction and tested the validity of those predictions by testing it on some of the data from the same WVR that wasn’t used for statistics. Why aren’t we doing that for climate models? That is, why aren’t we using the data from say anything before 1995 to make our models and then see if the models match what we observe from 1995 to the present? What is so difficult with that?

I asked a modeller that once, and that is one of the tests they do.

David Gould (Comment#30225) January 14th, 2010 at 6:14 pm queries “.. is tamino removing the worst models under his metric, or not?”

It’s hard to tell since he doesn’t have any quantitative metrics. It seems that it is all in the eye of the beholder …. in other words, how good the fit looks when graphed with very wide spans for both temp and time.

Charlie A,

Well, he specifically talks about data prior to 1960. His graph goes back to 1880. Graphing the other models data for that period, do any fall outside the line of the CCMA ones?

Can we classify Tamino’s time travel technique as being in a similar class with Rahmstorf’s? Rahmstorf travelled ahead in time to create his pretend numbers from the future. He smoothed his projected numbers with the real data and determined that the resulting average was higher. This supposedly proved global warming was “worse than we thought”.

Tamino travels back in time in order to project a future that is now past. Damn, climate science can give you whiplash.

Ok Tamino does have quick line at the end of his post that the comparison continues to be favourable even in the 21st century which I had missed on first reading. On my previous reading I only saw Tamino comparing model predictions against observations for the past.

If you want to compare Monkton to Tamino:

Monkton shows only his revised version of the IPCC projections against observations, and the result is radically different if you compare the true IPCC projections against observations.

Tamino compares both the true IPCC projections and his revised set against observations. I would encourage readers to have a look for themselves and form their own opinion on whether the difference between IPCC original and Tamino’s revised version is significant enough to warrant any accusations of trickery.

On a technicality, Tamino may have been careless in stating that his revised predictions put the lie to claims that recent observations somehow ‘falsify’ IPCC model results. However his first comparison is of the true IPCC model results with observations and this does put the lie to such claims…

David Gould:

However, tamino is looking at a longer model output/observations comparison than that. So, over the period that he looks at – 1880 onwards – are there any worse models than CCMA? In other words, is tamino removing the worst models under his metric, or not?

I think the problem some people are having here is this statement of Tamino’s:

The outstanding agreement holds not just for the 20th century, but into the 21st as well — putting the lie to claims that recent observations somehow “falsify” IPCC model results

Leave out that statement and it’s a decent analysis, IMO.

Tamino keeps making this claim by the way:

The biggest disagreement is just prior to and during world war II, when the method of measuring sea surface temperature changed, which may have caused a discrepancy in the observed temperature data

That doesn’t explain why the land data also exhibits this anomaly.

By the way, what do you guys make of this?

The El Niño-Southern Oscillation Phenomenon (Introduction to Religion) I”m pretty sure “Introduction to Religion” isn’t actually a subtitle to the book

Regarding Tamino’s analysis “observations somehow “falsify” IPCC model results”. If what Tamino is saying is correct, he has essentially pointed out an error in the IPCC model results himself, hasn’t he?

David–
The 20th century runs don’t go back to 1880– maybe one or two do, but most start sometime in the 20th century. So it’s impossible to compare back to 1880.

However by my reckoning the model with the worst performance in the 20th century– starting comparison in 1905 is bccr_bcm2_0. ratio of the Hadley rise to the bccr rise is 2.351. The ratio of the BCCM rise to the observed Hadley rise is 1.859.

However, if you don’t do ratios, the absolute differences are -0.0038 C/year vs 0.0057 c/year making BCCM worse in this sense.

Now, in my opinion, if all your are going to do is compare trends, I think ratios can make sense. The reason is the lower end is bounded to a large extent by zero– that is the trend we expect for any and all climate sensitivities with no forcing. If no forcings are applied expect simulations should show a trend of 0 C/century on average even if a model is totally, utterly and completely wrong. In contrast, if some positive forcing is applied to a model, models with small sensitivity will rise a small amount, and those with infinite sensitivities can rise an infinite amount.

If one wants to do something fancier based on honest to goodness statistical analysis, then we can bag the ratio issue. But, in that case, someone should do the hoest to goodness statistical analysis, apply it to all the model runs and apply in some rational way.

lucia,

Thanks. However, that seems to suggest that *if* tamino is using absolute differences then he is not being inconsistent in his methodology (ie, he is not arbitrarily excluding BCCM – although I guess that depends on where we draw the line at ‘arbitrarily’).

lucia,

An off-topic question: how exactly does one work out autoregression coefficient when dealing with AR(1) data? I have been able to find some stuff on this, and it seems to be just the slope between the data set X1 to Xn-1 and the dataset X2 to Xn. Is this right, though?

And is this autoregression coefficient an exact value or does it have some kind of range that can be estimated?

David Gould,
Yes, it’s the slope of y = e[t] and x = e[t-1]. You can easily calculate it without the need for regression with sum( e[t]*e[t-1] )/sum(e^2).

Lucia,

The 20th century runs don’t go back to 1880– maybe one or two do, but most start sometime in the 20th century. So it’s impossible to compare back to 1880.

The 20C3M runs begin between 1850-1900.

Chad,

Thanks.

David Gould (Comment#30205) It is impossible to be unfair eniugh to Tamino. He is the exemplar of a that can be wrong in Blogging.

If historical temperature data for certain periods in the past (eg 1930′s) keeps changing/drifting relative to recent data, how can we determine the fitness of any model?

Tamino’s work is a great example of what happens when faith is bolstered by rationalization instead of reason.

Just like the system-based gambler who always realises that his system will NOW work, because he his latest losses have shown him the adjustements he SHOULD have had n the first place.

Tamino needs a week on a quant trading desk to see the folly of his ways – and relearn some basics of probability.

David–

Thanks. However, that seems to suggest that *if* tamino is using absolute differences then he is not being inconsistent in his methodology (ie, he is not arbitrarily excluding BCCM – although I guess that depends on where we draw the line at ‘arbitrarily’).

Sure. He developed the method of exclusion after observations showed that method would have worked had it been applied before the observations came in. But before the predicted values were observed, there would be no particular reason to pick absolute differences instead of how far off by %.

But when you said he excluded the “worst”, BCCM is not objectively worse in the hindcast. It’s only “worse” if you can pick your method of diagnosing worse after your overall predicitons are shown to be wrong.

Chad–
The Climate Explorer doesn’t seem to have runs back to 1850 for many of the cases. Did you get that from PCMDI? The earliest data that I hvae for all runs is 1905.

David Gould–
Chad answered you while I was asleep.

The first lag correlation is also easily computed using corel(x,y) in excel and that function will be preprogrammed in any statsitical package. It’s just the ordinary correlation but for “x”, and “y” being the same thing lagged by one time step. You can find discussions of autocorrelation in books on statistics, but also physical sciences.

The autocorrelation (which is the graph of the lagged correlations at all time separations) is discussed in textbooks on turbulence a lot. Interestingly, the classic corrections for the ratio of numbers of samples to effective numbers of samples N/Neff would be called the ratio of the actual lapsed time to the ‘intergral time scale’ in a book on turbulence.)

Chad–
My file for miroc3_2_hires starts in 1900. I’m looking for the one that must start in 1905. (My program to re-baseline just looks for the last start date and prints out from that, but I don’t want to click them all open to see.)

Do you have data for miroc hires earlier than 1900?

Michael Hauber

On a technicality, Tamino may have been careless in stating that his revised predictions put the lie to claims that recent observations somehow ‘falsify’ IPCC model results. However his first comparison is of the true IPCC model results with observations and this does put the lie to such claims…

How does his first comparison do that? The projections– that is parts going into the 21st century are off. A graph of Temperature V. Time, rebaselined to force agreement during 1980-2000 masks the disagreement and cannot put lie to any such claim.

The Temp. V. time graphs for the multi-model mean hindcast were never wildly off. This has never had anything to do with how we diagnose the ability to forcast or project. Arrival of future data cannot change the agreement between the similations and the data in the hindcast.

Lucia,

Do you have data for miroc hires earlier than 1900?

No. MIROC HIRES starts in 1900 for me to. Yes, this is PCMDI data.

Here are the start dates I have:

model run start.date
BCCR BCM 2.0 1 1850
CCCMA CGCM 3. 1 1850
CCCMA CGCM 3. 2 1850
CCCMA CGCM 3. 3 1850
CCCMA CGCM 3. 4 1850
CCCMA CGCM 3. 5 1850
CCCMA CGCM 3. 1 1850
CNRM CM 3 1 1860
CSIRO MK 3.0 1 1871
CSIRO MK 3.5 1 1871
GFDL CM 2.0 1 1861
GFDL CM 2.1 1 1861
GISS AOM 1 1850
GISS AOM 2 1850
GISS EH 2 1 1850
GISS EH 2 2 1850
GISS EH 2 3 1850
GISS EH 1 1880
GISS EH 2 1880
GISS EH 3 1880
GISS ER 1 1880
GISS ER 2 1880
GISS ER 3 1880
GISS ER 4 1880
GISS ER 5 1880
IAP FGOALS 1.0g 1 1850
IAP FGOALS 1.0g 2 1850
IAP FGOALS 1.0g 3 1850
INGV ECHAM 4 1 1870
INM CM 3.0 1 1871
IPSL CM 4 1 1860
MIROC 3.2 HIRES 1 1900
MIROC 3.2 MED 1 1850
MIROC 3.2 MED 2 1850
MIROC 3.2 MED 3 1850
MIUB ECHO G 1 1860
MIUB ECHO G 2 1860
MIUB ECHO G 3 1860
MPI CGCM 2.3. 1 1851
MPI CGCM 2.3. 2 1851
MPI CGCM 2.3. 3 1851
MPI CGCM 2.3. 4 1851
MPI CGCM 2.3. 5 1851
MPI ECHAM 5 1 1860
MPI ECHAM 5 2 1860
MPI ECHAM 5 3 1860
MPI ECHAM 5 4 1860
NCAR CCSM 3.0 1 1870
NCAR CCSM 3.0 2 1870
NCAR CCSM 3.0 3 1870
NCAR CCSM 3.0 5 1870
NCAR CCSM 3.0 6 1870
NCAR CCSM 3.0 7 1870
NCAR CCSM 3.0 9 1870
NCAR PCM 1 1 1890
NCAR PCM 1 2 1890
NCAR PCM 1 3 1890
NCAR PCM 1 4 1890
UKMO HAD CM 3 1 1860
UKMO HADGEM 1 1 1860

Looks like the formatting didn’t survive. But it’s easy to read nonetheless.

Hhmmm… I’m going to have to look at why a particular compilation I have started in 1905. I thought I had the program scan for the earliest valid year and then spit out from there. Maybe I just requested starting in 1900.

Still… in terms of David question, I can’t really answer which has the worst trend starting in 1880 because a number of models don’t go back that far. So, I answered on the worst trends since 1905. That happened to be the start date in one of my compilations files.

Thanks Chad– I also figured out why the file I looked at started in 1900 instead of 1905. I could redo my answer for David, but I don’t think the 5 year difference will matter to what either of us think.

Lucia,
Yeah, what’s five years compared to a century?

Lucia:
“and whether or not you correct for ENSO ”

Do you have an ENSO correction algorithm, Lucia?

Tilo–
I’m using a multilinear regression accounting for ENSO using MEI. I haven’t discussed that– but it’s straightforward.

Funny Lucia,

When I suggested that we eliminate models that didnt hindcast well
the AGWers went bonkers ( lots of bad models = wide predictions)

When I noted that they winnowed models ( pick good ones) to do
attributions studies ( models without drift = narrow predictions)
They said this was good proceddure.

Need narrow CIs to rule out natural variation: winnow models
Need wide CI to have a hard to falsify forecast: Democracy of models. Give me your tired, your unskilled, everyman models

Look

I suggested that models be winnowed by handcast performance and I could not get gavin to agree that this was a good proceedure.

This was before things dipped cool

Steven–
If I’m not mistaken, you repeated that mantra over and over. Roger Pielke Jr. has mentioned the issue of modelers giving a huge range or predictions and declaring victory by discovering that, having said absolutely anything might happen, discovering that absolutely anything did.

What is interesting is Tamino essentially decreeing that one model looked false in the hindcast, still looked bad, and then decreeing that if we treat that model as false, we can’t say we found the models false!

Yes,

you know it might be interesting to rate the models in terms of their hindcasting capability.

Then test the forecasting ability by inlcuing increasing number of models.

With a small number of “good” models you will get a wide spread because of small N.. so you increase the models….I think There might be a trade off or at least a curve that shows hindcast error
versus forecast CI.. get what I’m angling at

steven mosher (Comment#30499),

I think it is important to point out that “good” and “bad” models can’t be so declared based only on the hindcast. A truly “good” model hindcasts well AND has the right climate sensitivity. Heck, we know that they are all “tuned” (AKA optimized, AKA “fitted” to the historical data), so hindcast performance insn’t much of a constraint. All models that are truly correct have to predict the same climate sensitivity. They do not. So who knowns which (if any) is good.

Everyone is missing the point on climate models. They can only replicate PAST climate. They have been calibrated on the solid records available from the most impeccable sources, CRU, GISS and CDRC. Their accuracy is impeccable in predicting known results. But as Yogi Berra said, “predicting is hard, especially about the future.”

I just went through a review of the 9 pages of references, a total of 537 papers, on Chapter 9, Understanding and Attributing Climate Change, of the IPCC 4th Assessment Report AR4. I did not read all of the papers. From the titles, I selected those that appeared to address attribution. By searching the web by the title, I was able to find the abstract on nearly all of the papers and the complete PDF on most of the papers. I COULD NOT FIND ANY PAPERS DESCRIBING ANY EMPIRICAL EVIDENCE OF ANTHROPOGENIC GLOBAL WARMING (AGW). To put it in context, it reminded me of a debate among theologians about how many angels can dance on the head of a pin. The evidence behind AGW amounts the Argument from Ignorance, the models using (their definition of) natural causes alone cannot replicate the observed global warming; therefore, it must be caused by humans. Now all of a sudden the AGW alarmists have discovered NATURAL VARIABILITY and that is the cause of the cooling. I have a post on my website on how you can calculate natural variability. It is interesting to see that all of the global average annual temperatures since 1880 using NOAA National Climatic Data Center data, after all of their adjustments, still fall within plus or minus 3 standard deviations, which is an accepted measure of natural variability.

http://www.socratesparadox.com

Lucia,

It might be interesting to do the opposite of Tammy.
Pick a model that gets it wrong by being too “cool” toss that one out

Bruce,

AGW is argued with every stupidity the human mind can fathom. Just check out THE DEVIL thread. What a colossal waste.

But I guess what they say is true. Once a person crosses the threshold into Brainwashdom, it’s hard to get ‘em back. It’s like Toyland. Once you pass the border, you can’t return. It’s the Hotel California. You can check out anytime you like, but you can never leave.

For the AGW’ers, religion, politics and science are all blended together into a person’s own Religion of Political Scientism. And so the beat goes on… I suppose we must, too. (sigh)

Andrew

Bruce, there are reasons climate models don’t do a great job with forecasts that don’t affect their utility for determining CO2 climate sensitivity.

Andrew_KY, that word “blind” fits you pretty well.

Bruce–
I agree with Carrick. The sensitivity could be right and the ability to forecast could still be terrible. There are plenty of other things that can go wrong.

Carrick,

About the “blind” comment,

When I see your AGW brethren using:

1. Letters of The Alphabet
2. Polar Bears
3. Squggly Lines
4. After-The-Fact Adjustments
5. Scaremongering
6. Grandiose Assertions
7. Silence
8. Models That Don’t Predict Anything
9. Papers They Don’t Understand
10. Links That Don’t Work
11. Warm Weather Events (Isn’t that just weather?)
12. Data They Have No Idea Wether It’s Right or Wrong
13. Excuses For Not Providing Information
14. Novel/Dubious Statistical Methods
15. Mysterious and Magical Computer Code That Provides The Right Answer
16. “Educate Yourself, Andrew” instead of “This Is What I Know, Andrew: ”
17. “I’m a Scientist”
18. You Can’t Ask Those Questions
19. You Can’t Talk About That
20. You Are A History Revisionist/Denialist/Creationist/Right Wing Fanatic

And that’s just the Top 20.

Am I “blind” when I see these things?

Andrew

Andrew_KY, they aren’t my “brethren” except in the same sense you are. Yes, I find a lot of what they do objectionable, but I’m no more responsible for their behavior than you are.

That’s as far as I care to go with this.

carrick seems like a person I can have a rational discussion with