in my experience, a few hair-raising Monte Carlo exercises with absolutely random data sets usually knocks the “pretest sillies” out of almost every student, at least as far as what they know goes. Whether they will resist the temptation to sin is quite another story. The publication pressure gets worse every year, and there’s almost nothing easier to hide than pretesting. Personally, I think a lot of young scholars get very bad incentives (quantity versus quality) and it gets worse all the time.

But you need to have an objective, pre-hoc definition, not post-hoc analysis.

It’s amazing how straightforward this concept is for non-climate scientists, while climate scientists don’t get it at all.

I am really struck by the argument that the “statistical properties” of the series can determine which site you choose. This seems difficult for me; if you are choosing the proxy as a measure of temperature, it is their relationship with temperature which is all important, and the temperature record will have its own idiosyncratic statistical properties (which you don’t know). The issue is whether the proxy informs about temperature, not statistical property.

Likewise, tree records must truly be informative about temperature. If some go randomly up, some go randomly down, choosing a few series which go in the same random direction as temperature over a short period isn’t necessarily going to be informative elsewhere. I can see this issue could be difficult, because you could argue that some tree series are temperature-sensitive, while others are not. But you need to have an objective, pre-hoc definition, not post-hoc analysis.

Certainly, choosing one site over the other seems to be arbitrary. But there is at least as big a can of worms in trying to resolve why there is so much difference between the two tree series; and this must surely open to question whether the series are valid temperature proxies.

cheers
per

Rob Wilson has written me offline, attempting to justify the switch on the basis that the variance of the Yamal chronology is more stable than the variance of the updated Polar Urals chronology.

First, since you’d sent me this argument offline and because you’ve always been cordial, I didn’t want to present the graphic, but I’ll post it up to better represent your argument.

Second, I don’t necessarily agree with your statistical arguments. If you’re relying on some fairly subtle statistical arguments to justify a substitution that, as you agree, has a significant impact on the final results, then it would be nice if these methods were actually mentioned in the article or its SI. In this case, they weren’t. However, I’m happy to present your side of the argument.

In this particular post, I make no reference to decisions made in D’Arrigo et al 2006, but limit my comments entirely to Briffa’s substitution. There’s no evidence that Briffa relied on a statistical analysis of the type that you carried out.

I apologize if the net result of these posts has not accurately reflected your position; I usually bend over backwards, verging on pedantic, in my statement of positions that I’m criticizing, including detailed quotes and actual graphics. The only reason for not doing so here was that you’d sent me the graphic offline, but I’ll redress the situation forthwith. Cheers, Steve

For others, I wish to emphasize that Rob Wilson is an earnest and serious researcher, who, aside from that, has been consistently civil and gracious with his time with me. However, he did not make all the decisions in respect to D’Arrigo et al 2006. I think that I am entitled to criticize those decisions as eventually published. Readers should understand that Rob disagreed with some decisions (e.g. archiving) and keep that in mind in my discussions of this article.

Help?

Also, Rob, thanks for posting, but I fear I didn’t understand your post.

First, if you couldn’t get the raw data … couldn’t that be construed as a clue as to whether you should include the processed results of that mystery data in a scientific paper? It makes the study unreplicatable …

Second, why was the Yamal data-set “optimal”? You mention it is for “clear statistical reasons” … but since as you say, you could not get the raw data, how on earth did you obtain the clear statistics?

w.