Via Richard Klein's Twitter feed comes this interview with Raymond Bradley in which he discusses his new book. Fascinating stuff, particularly this bit:
In 1998, a post-doc, Mike Mann, Malcolm Hughes and I published an article in Nature on climate in the last 600 years (Mann et al. 1998). Then, in 1999, we published another article in Geophysical Research Letters on temperature over the last 1000 years (Mann et al. 1999). The title was “Northern hemisphere temperatures during the past millennium: inferences, uncertainties, and limitations.” We were emphasising the uncertain nature of the problem. But nevertheless, when it got picked up by the summary for policymakers of the third Intergovernmental Panel on Climate Change (IPCC) report, important caveats were left out.
Now who was it who was the lead author on the paleoclimate chapter of the Third Assessment Report again? I will need to revisit the treatment of the Hockey Stick in the report and the summary for policymakers to see just where the caveats were missing.
And then this:
[In] the IPCC Third Assessment, a report of over 880 pages, the “hockey stick” occupied less than one page. There were more than 200 figures in the book. The “hockey stick” figure was only one of them.
Six of them, if I'm not mistaken.
And then there's this:
Besides, we have since shown that even if you entirely avoid the procedure Barton’s statisticians objected to, and simply average all the data we used, you get the same “hockey stick” result. Simply put, the “hockey stick” is bomb-proof. No amount of data manipulation will make it go away.
This is very icky. The reason principal components analysis was used was to summarise down the US tree rings so that they weren't overrepresented in the dataset. If you don't summarise them down then they are indeed overrepresented. It's still wrong! And he can't get away with the fact that the hockey stick shape is coming from the bristlecones and everyone agrees that the bristlecones are contaminated with a non-climatic signal. Putting forward a different way of processing contaminated data is not going to convince anyone.