Betting

The Blackboard Projection of Official UAH Anomaly.

34 Comments

Drumroll please: I have several announcements to make about The Blackboard Commenters Prediction of the Official Observed UAH Lower Tropospheric Temperature Anomaly for July.

  1. The UAH LTT anomaly for July was 0.410 C (as reported by Dr. Roy.)
  2. The Blackboard multi-commenter sample mean projection based on an ensemble of 21 participants was 0.366 C which was 0.044 C below the observed value.
  3. I applied the “does the official value fall within the distribution” test to determine whether the projections of a multi-commenter ensemble failed to bound the observations. (This appears to be the test currently favored by “some” modelers. Diagnosis: Our method did not fail. Therefor, those who like this test should decree that there is no evidence this method does not work.

    For specifics: The standard deviation of our guesses was 0.101. Assuming the guesses are drawn from a population of all possible guesses and that population is normally distributed, the 95% confidence interval for that population is estimated to be ±0.215C, resulting in a range from 0.264C to 0.685C. The observed value of 0.410C falls well within this range.

    To relax the assumption that guesses are normally distributed, I observe that 40% of the commenters guessed too high while 60% guessed two low. So, the true value falls well within the distributions of projections.

  4. I applied a t-test of the sort on learns as a sophomore in college applying no correction for any autocorrelation in the projected values. This tests whether the multi-commenter mean that we would obtain if we drew an infinite number of samples is consistent with the observed value of 0.410C. No uncertainty is attributed to Roy’s observation.

    Applying this test, I determined the standard error by dividing the square root of 20 (i.e. 21 participants-1). The standard error was found to be ±0.022C. The multiplier associated with the 95% confidence interval was 2.093; therefore the 95% confidence intervals was estimated to be ±0.046C. This means any observed value between 0.319C and 0.412 C is “not inconsistent with” the multi-commeter mean projection.

    Roy reported 0.410C! (Whoo hoo! Our method of predicting the multi-commenter mean doesn’t fail the test!)

  5. I took the extra step of noting that 8 out of 20 commenters guessed too high. Assuming the probability of a commenter guessing too high is 50%, and the hypothesis that the true- multi-commenter mean equals the observed values, I computed the probability that 9 or fewer commenters would guess high. I found a 33.3% chance this would occur. So, using this tests, the multi-commenter mean projection is not inconsistent with Roy’s official observation of the LTT for July.
  6. Ryan O had the closest estimate. His projection of 0.411C exceeded Roy’s official value by only 0.001C.

The remaining projections are tabulated below.

Commenter: Projection Error Abs(error)
RyanO 0.411  0.001 0.001
Terry 0.404 -0.006 0.006
Douggerel 0.416  0.006 0.006
John F. Pittman 0.418  0.008 0.008
Andrew_FL 0.400 -0.010 0.010
duwayne 0.400 -0.010 0.010
Steve F 0.421  0.011 0.011
Chuck L 0.397 -0.013 0.013
Zer0th 0.392 -0.018 0.018
Lucia 0.435  0.025 0.025
George Tobin 0.445  0.035 0.035
David Gould 0.472  0.062 0.062
Zeke 0.485  0.075 0.075
Martin 0.321 -0.089 0.089
VG 0.500  0.090 0.090
DG 0.299 -0.111 0.111
Greg Meurer 0.294 -0.116 0.116
Basil 0.210 -0.200 0.200
Bob Tisdale 0.205 -0.205 0.205
Bill Illis 0.202 -0.208 0.208
BBuckner 0.150 -0.260 0.260

We can give this another whirl next week. For those hoping this game can last forever: it will have to be modified at some point. Dr. Roy posted this:

NOTE: For those who are monitoring the daily progress of global-average temperatures here, we will be switching from NOAA-15 to Aqua AMSU in the next few weeks, which will provide more accurate tracking on a daily basis. We will be including both our lower troposphere (LT) and mid-tropospheric (MT) pre-processing of the data.

When that occurs, the daily measurements will provide too much information to let anyone call “projections” made on the final day be projections of any sort. So, we’ll either have to switch to guessing RSS based on the Aqua-AMSU or posting our projections earlier.

34 thoughts on “The Blackboard Projection of Official UAH Anomaly.”

  1. Lucia, so we can say that The Blackboard Commenters Prediction of the Official Observed UAH Lower Tropospheric Temperature Anomaly (TBCPCOULTTA) is a robust method for short term prediction of the UAH monthlies?

    We might also say that RyanO may very well be a ringer if he continues his accurate predictions. The first two letters of Roy are also the two capitalized letters in his sign-on. Think about it. Dunno! Just sayin!

  2. CoRev–
    I don’t think we can call it ‘robust’ until we fail to fail to project twice.

    Failing to fail to be right at least twice seems to be the criterion for “robust”. Once that has occurred, failing sometimes does not seem sufficient to decrees a method previously thought “robust” is “not robust”.

    Robustness of a method of projecting things becomes the null hypothesis, and we must show it fails over and over by any and all possible criteria.

    I’m pretty sure this is how “robust” works.

  5. I’ll bet that if you tried correlating the guessed values with the guesser’s personal evaluation of AGW it might explain the rather odd distribution of guessed values. A “Humans are destroying the Earth!” warmer like Jim Hansen would be a 3 or 4 DPD (degrees per doubling), while a “You have got to be joking!” denier like Dick Lindzen would be a 0.5 DPD, or maybe even a 0 DPD.

    Once the DPD effect is determined (probably a linear effect of DPD on guess, but you never know), it can be used to adjust the guesses to their “correct” values. I believe the adjusted distribution of guesses would be normal, and the mean guess close to the correct value.

  6. Lucia, I am a little concerned that in your analysis of my response you chose to not comment on my RyanO/ROy analysis. Some what troubling, me thinks. What do you know???? Come clean!

    I accept your understanding of “robustness”, and think it far superior than others I have nearly almost seen in other writings.

  7. No, Lucia, Douggerell is in 3rd. He was tied for 2nd but is already on record as giving up the battle for 2nd and has accepted the third place on the podium.
    The Awards Protocol Committee is much relieved that the standard Awards Ceremony can be used.

  8. There is no evidence that the orchestra failed to play the correct national anthems during the Awards Ceremony.

  9. Scooter,
    I believe the national anthem of Wisconsin is “We Love Cheese”. The melody is by Handle. (You probably have heard the lyrics at Christmas. The full chorus sings “Oh we love Cheese. We love Cheese. And all our cheese is gra-ai-ai-ai ted. Is grated. Is grated….” and so on. This group is singing it. It’s difficult to understand many of the lyrics

  10. Congratulations to the top five places. This was an difficult projection. Next three months, though, will be much more challenging and will prove to separate the “men from the boys”.

  11. Lucia,

    Does the Wisconsin anthem refer to cow’s milk cheese or sheep’s milk cheese? I couldn’t quite make out all the words. Did Handle ever try Wisconsin cheese, or did he write based only on its reputation for quality?

  13. Actually my guess was .123 after hansenizing, Lucia. Hope that doesn’t throw a stick in your calculations.

  14. Martin–
    Hansen isn’t involved in UAH, so I took your UnHansenized projection for UAH. Hansenized projections apply to GISS.

    SteveF– Handel traveled in high society. I’m sure he tasted Wisconsin cheese. Douggrel can tell us whether the anthem refers to cow or sheep’s cheese. It’s always so difficult to make out the lyrics when listening to that anthem.

  17. Meh, tied for second, third, whatever. I was over, Terry was under. Considering I chose that number by counting the number of cows I can see out my window (less than one half of a cow; there was rounding involved), I’m happy enough.

    Sheep’s milk cheese? What are you smoking? This is WISCONSIN. Did you at least catch the bratwurst reference in the video?

  18. I would suggest that many of us would consider other factors besides the raw statistics of hypothesis testing in assessing whether this was a good prediction. We could consider other factors such as the ‘internet polls are always wrong’ heuristic.

    Or consider whether the participants used reasonable methods and data to get to their predictions. If we consider the ‘count the number of cows outside my window’ is typical of the methods used we may guess that the poll was only accurate due to luck. If we consider most people made a reasonable assessment of the evidence (the daily Uah trend), and that the daily Uah trend was reasonably indicative of the monthly trend, then we might conclude that the poll was accurate for good reasons.

  19. The problem is that the projection was a mish-mash of different methodologies (mine I can’t quite remember but it involved averages and assumptions and in a way assumed some “warming”-since I got an answer I kinda liked, I didn’t throw it out!) some valid others obvious guess work. Some were obviously drawn towards Lucia’s estimate while at the same time having totally bogus methods of getting their particular results. Ryan O’s method seemed to be to note that the anomaly would probably be close to .4 something and decided to call information ;)-he ended up getting the closest, but not entirely by serendipity, only partly. Others just made stuff up or were impacted by group think (note some clustering towards .4 something).

    This vaguely reminds me of that monkeys thread a while back.

    Here’s a question no one seems to have noticed: the anomaly from June jumped up by .002 from last month-what’s with climate data and data set creep?

  20. Andrew_FL,

    It did not jump as far as I can see. The .003 figure is *smaller* than .01.

  21. Based on RyanO’s track record, my meta-prediction is that the official UAH Anomaly will soon be revised upward 0.001C to match RyanO’s prediction.

  22. The July UAH increase was the biggest monthly change in the entire UAH record (in terms of the increase and in absolute terms).

    The next closest global monthly change was 0.293C so this last month has to be considered very unusual.

    The southern hemisphere change at +0.634C is nearly double the previous record (I have the southern hemisphere records going back to 1850 and the southern hemisphere over the longer record is known for having very wild swings and is very unlike the more normal temp record for the northern hemisphere).

    The change in the tropics in July 2009 at +0.434C is the second largest ever and the biggest one was (maybe expected), July 1997 – just as the Super El Nino was really taking off – +1.85C anomaly in Nino 3.4 in 1997 versus +0.88C this year.

    So overall, something really strange happened this month and it is tied into the record warm temps that Antarctica is experiencing.

  24. Andrew_FL,

    Hmmm. I go by the vortex site, so that is where my error stems from.

  25. More fun than fantasy football, almost as fun as fantasy baseball. Congrats, Ryan – ya $#@!! @#%$ 😉

  26. Terry,

    I got dibs on marketing Global Warming: The Gameâ„¢

    People are already playing. 😉

    Andrew

  27. David-the correct vortex site for comparing to Roy’s monthly posts is the one which gives the anomaly out to three decimal places, not two:
    http://www.nsstc.uah.edu/data/msu/t2lt/tltglhmam_5.2

    I would have linked there, but soon it will change (by the way, I don’t know why the values in that file are so different from the others-go figure!)

  31. Uh Oh, Ryan responded and claims not to be Roy. Maybe I was wrong. I do know this though, next contest I will play. Why??? Because I too want to be a winner and associated with the other winners in the Global Warming is all about teh… umh, uh. Let me get back to Y’all.

    Which leads me to this question. If the temps can change at this rate in such short periods, can anyone relate CO2 as a “direct” cause? If so, how?

  32. CoRev

    Which leads me to this question. If the temps can change at this rate in such short periods, can anyone relate CO2 as a “direct” cause? If so, how?

    I’m glad you are going to participate in the next pool. I’ve written a script to make it easier for everyone to log their bets. (I won’t have to look through comments.)

  33. Lucia, thanks for the script. I am sure my model will be the best of all, well almost, anyway. If anyone else is thinking of just copying RyanO’s then you will just be copying my own method (Ummm model.) Of course that model could change month to month depending on the previous month’s result.

    Lucia, on the serious side, is it just me who thinks the recent rapid changes takes the steam out of the AGW argument?

  34. In case anyone is interested in my methodology:
    .
    Step 1: Pick a word that has a numerical representation in slang.
    .
    Step 2: Use the numerical representation as your guess.
    .
    “411”=slang for “information”.
    .
    This, incidentally, is why I am positive Roy had a rounding error. My method is foolproof.

Comments are closed.