Wegman Wiggles Out of Hockey Stick Accountability

Here's a great RealClimate.org post on the hateful hockey stick. On one hand, I hesitate even to raise the subject because deniers have used the famous Mann hockey stick graph so effectively to distract attention from the scientific consensus on climate change. On the other hand, it now seems that one of the most recent and most-quoted hockey stick critics. Dr. Edward J. Wegman, is dodging his responsibility to back up his numbers.
Also, if you read down on the RealClimate post to comment #6, you'll find an interesting bit of criticism of the graphs that Wegman uses to make his critical argument. And if you click through to Wegman's voluminous report, and go to figure 4.1 on page 30, you'll find the graphs in question originated not from Wegman, but from McIntyre and McKitrick, the economist and the statistician who first criticized Mann's original work.
Which leads back to the good question: if they were truly interested in launching a fair critique of the Mann graph, why wouldn't they keep the y-axis the same in comparing Mann's work with theirs?
It summons to mind the old aphorism about lies, damn lies and statisticians.


I think you are missing something here. What Wegman wanted to show was that by using the Mann et al method, you could get the same result that Mann et al got by feeding in not data, but noise. In other words, the Mann et al method data mined for the hockey stick shape.

Although Mann et al could have done their work using noise (it would have produced the same graph), it would hardly have produced a paper with the same impact as MBH98.

In other words, the Mann et al method data mined for the hockey stick shape. (This reply is months late – I just stumbled on this while perusing desmogblog.com – but I’m posting this reply to make sure this information gets on the record here) There is a huge difference between Mann’s hockey stick and McIntyre’s “noise-only” hockey stick. How do you know that they are so different? You look at the eigenvalues. The eigenvalue associated with Mann’s “hockey-stick” leading PC is much greater than the eigenvalues associated with the “hockey-stick” leading PC’s that McIntyre generated from random noise. When you perform a PC decomposition on a data set, the first thing you need to do, before you do anything else, is to look at the eigenvalue magnitudes. A quick look at the eigenvalues would immediately show that McIntyre’s “hockey-sticks” were insignificant (representing a tiny portion of the data), while Mann’s “hockey stick” truly was a dominant feature of the data. The fact that McIntyre didn’t bother to look at eigenvalue magnitudes means that he is either (1) incompetent, or (2) dishonest. If you don’t know what an eigenvalue is, you have no business participating in this discussion.

(this reply is months late - just stumbled on this myself - but appropriately and ironically only a few days after Caerbannog) Caerbannog, McIntyre would probably get your Monty Python reference and get a kick out of it. As far as your issue with eigenvalues go, take it up over at Climate Audit. There’s a search function for the site that will pull up any number of threads discussing eigenvalues and eigenvectors. You can also have it up with McIntyre himself - he’s quite good with linear algebra (as are some of the regular posters).