Wednesday, May 02, 2007

Zero crossing....

James Annan has heard from his eagle friends about further goings with Hegerl et al. Annan has heard from Tapil Schneider about how

Hegerl et al used a regression to estimate past temperatures anomalies as a function of proxy data, and estimated the uncertainty in reconstructed temperature as being entirely due to the uncertainty in the regression coefficient. The problem with this manifests itself most clearly when the tree ring anomaly is zero, as in this event the uncertainty in the reconstructed temperature is also zero! Steve McIntyre's plot of the data that Hegerl et al supplied as supplementary information illustrates the effect neatly (and the problem is quickly diagnosed in the comments following).
Eli thought he had seen this before, and indeed he had,



in the discussion about Huang, Pollack and Shen's's boring hole long term global temperature reconstructions that were later commented on by the Fuzzy Blogger's Group Blogfather, Stoat. Stoat got an explanation from Pollack on the differences between different versions of their reconstructions which casts considerable doubt about the one above. Of course, the one above also was a favorite of Lord help us Monckton which is what made us twitch our noses in the first place when he blew the first 1K year section up.


1600 is clearly a point of zero anomalie. Huang, Pollack and Shen do not quite have "isobestic points" in their confidence limits, but close enough to imagine the same mistake has been made. Monckton is, of course, a point of infinite error.

8 comments:

bigcitylib said...

And what does this mean for those of us who percieve your graphs as merely pretty lines on a paper?

What was the gist of Hegerl et al's reserach?

Anonymous said...

Pielke is out advertising another study on hurricanes/climate change. Gee, I wonder if this paper says that there is a link or that there isn't?

Oh, well it's a paper by Chris Landsea. The favorite hurricane researcher of the Bush Administration. So, no surprise at the results.

And no surprise that Pielke is flacking the study.

Mus musculus anonymouse

EliRabett said...

BCL, it is kind of twisty technical. Annan, Schneider and Hegerl were talking about estimating confidence limits (essentially error limits) for proxy temperature reconstructions (the sort of thing that Mann et al started). In the example given by Hegerl, at the point where the temperature anomaly (the difference between the measured temperature and the average temperature) was zero the error limits were also zero. If you look at the second graph you see that the same thing happens there at 1600. It is a little less clear in the first graph.

This is stupic. You don't expect the error to go to zero, simply because the measurement is zero. Schneider points out that this is a blunder on Hegerl's part. Annan, always fond of Hegerl's blunders piles on. Eli shows that this appears to be fairly common by pointing to Huang et al.

James Annan said...

Rather than portraying it as a feud, let's just say I was prompted by a correspondent who found my old comment and thought I might have been a little unfair in my peremptory dismissal of his complaint last time round.

I wouldn't like to be too hard on Gabi, I think it's great that she has managed to go so far towards alleviating my concerns over the AR4 (she was lead author on Ch 9, see my post on that topic). It must be a tough job when there are people saying diametrically opposing things, and it's hard to accept that relative outsiders such as I can actually be correct. The fact is that the first draft was a disaster and since then I've been struggling to save them from embarrassing themselves in public. I don't expect thanks for the modest measure of success I've had.

EliRabett said...

Sometimes it is difficult for Eli to control the snark and he has been known on occasion to go too far, especially given the opportunity to let you and him fight, however, there was a serious point lurking in the post

William M. Connolley said...

I'm not quite sure the situation with pert{HPS} is quite analogous... the lines there are not error limits, just different reconstructions... though why they should all be zero at 1600 unless its a tie-point I don't know

Anonymous said...

I wouldn't like to be too hard on Gabi, I think it's great that she has managed to go so far towards alleviating my concerns over the AR4 (she was lead author on Ch 9, see my post on that topic). It must be a tough job when there are people saying diametrically opposing things, and it's hard to accept that relative outsiders such as I can actually be correct. The fact is that the first draft was a disaster and since then I've been struggling to save them from embarrassing themselves in public. I don't expect thanks for the modest measure of success I've had.

Your vote of confidence in peer review and the abilities of IPCC authors to accomplish basic statistical tasks has been duly noted.

We shouldn't be so nasty and mean as to expect IPCC authors, whose pronouncements have consequences measured in the trillions of dollars, to be able to do a competent job of numerical analysis that undergrads are expected to do.

EliRabett said...

Cerberus, given the mathematical blunders that Lord McKitrick has made, Eli thinks those who guard glass houses would be more careful.

Actually this is a good example of how things work. A preliminary peer review and then other folk look at and evaluate the work.