Canada’s * National Post*, in a piece from a little more than a year ago, made a call for more statisticians to be involved in climate change research, much as the American Meteorological Society recently did. It’s relevant again, because the

*Post*article is highlighted today on the indispensable ClimateDebateDaily.com.

The title of the article is somewhat ridiculously called “The Deniers — Part I,” as if the only opposite course is to be devout and be a “Believer.” But let that pass. There are IX–excuse, me, 9—other parts to the series, highlighting topics like “warming has benefits” (true), “the sun moves climate change” (true), and “limited role for CO2” (also true). None of these are strictly “denialist” positions; they are, in fact, attempts to fully understand the physics of the climate system. To call those who study these areas “deniers” shows, then, how far we’ve slipped from sanity.

Anyway, Part I is more or less an interview with Ed Wegman, who is “professor at the Center for Computational Statistics at George Mason University, chair of the National Academy of Sciences’ Committee on Applied and Theoretical Statistics, and board member of the American Statistical Association.” Wegman was the guy who investigated Michael Mann’s “hockey stick” temperature curve and found the statistical methods behind its data analysis wanting.

Ed says that if “statistical methods are being used, then statisticians ought to be funded partners engaged in the research to insure as best we possibly can that the best quality science is being done, [and] there are a host of fundamental statistical questions that beg answers in understanding climate dynamics.”

One place to recruit these statisticians is from the American Meteorological Society’s Probability & Statistics Committee, but Ed is suspicious: “I believe it is amazing for a committee whose focus is on statistics and probability that of the nine members only two are also members of the American Statistical Association, the premier statistical association in the United States, and *one of those is a recent PhD with an assistant-professor appointment in a medical school*” (emphasis mine).

My readers, who are lovers of logic, will have instantly noticed that Ed has committed the “appealing to authority” fallacy when he implies that the poor schmuck stuck in the medical school cannot possibly know anything of climatology.

My friends, that schmuck is me. The other ASA member is Tilmann Gneiting, of the University of Washington, who is brilliant and one of the world’s biggest sweethearts. I can also set your minds at ease by telling you that the non-ASA members are no slouches at statistics, and they know a great deal of physics.

I wrote Ed to let him know that if he wanted to check my record, he might find that he and I are not too far apart. But Ed’s pretty busy and I’m still, at 1+ years, waiting for his response.

No, my point of writing this post was not just to stick it to Ed a little. It turns out the *Post* article and Wegman’s comments are topical, because starting tomorrow the AMS holds in annual meeting in New Orleans. The Probability & Statistics Committee will meet at this time, and also sponsor a four-day conference. I give my global “hurricanes have no increased” paper; Gneiting has a paper, and so do several other statisticians. I’ll be (trying to) write daily updates about major papers and so forth.

I’ll also let you know if I see Wegman haunting the halls.

I’m familiar with elementary statistics, and linear regression, but I know absolutely noting about

auto-correlations. Can you recommend any books covering this subject?

Correlation is mentioned in just about any stats book, explained more or less well. I have a draft of a 101 book that you can look at, linked on my resume page, but it’s far, far from complete.

But positive correlation only means that if you take, say, two measurements in time and the first is high, then there is a high probability the second will be high, too. If the first is low, then there is high probability the second will be low.

Old fashioned correlation can even be graphed on a scatter plot. For the x-axis, plot the first measurement, for the y-axis, the second. If you plot a bunch of measurements and find a straight line more or less connects all the points, then you can say the data is auto-correlated.

William,

I’ve been following the national post’s ‘The Deniers’ series. They had originally written 9 parts, but have continued right on and are up into about 36 parts now. The original series, for whatever reason, doesn’t have links to the continuing series. You can see the whole series here: http://network.nationalpost.com/np/blogs/posted/pages/climate-change-the-deniers.aspx

-Rich

I don’t think I made myself clear. I’m familiar with correlation, I’m referring to

“Auto-Correlation”, where you try to weed out

spurrious correlatons when there’s a time feedback.

http://landshape.org/enm/autocorrelation-in-gam-and-grasp/

“Autocorrelation describes correlation between a process, say Xt, at a different point of time Xs. The autocorrelation function can be depicted in a formula”

“If you use time series data in regression analysis, autocorrelation of residuals will be a problem area, since it will lead to an upward bias in the statistical significance of coefficient estimates. A Durbin Watson test can be used to detect the presence of autocorrelation. You can use Durbin?s h statistic, if your explanatory variables include a lagged dependent variable. To avoid autocorrelation related problems you may use differencing of data and lag structures in estimation.”

Any books covering the Durbin Watson test, or possibly more effective tests for auto correlaton?

– A. McIntire

Alan,

Almost any book on time series has the tests you are after. A good one is Brockwell and Davis, “An Introduction to Time Series.”

But it sounds like you’re almost there. Auto-correlation is the same as correlation, except it involves variables x_t and x_t-1 instead of x and y. As a simple test you can plot each x_t by each x_t-1, and if you can see some kind of signal in the data, then they are auto-correlated.

I am sure that a lot of people who visit this site would be very interested to read your views on the way Tamino deals with PCAs and the hockey stick on his recent posts on his “Open Mind” blog.

Wegman has been a strange actor in all this drama…