We mentioned last week a new paper by University of Guelph professor Ross McKitrick arguing that the way IPCC scientists have been looking for “fingerprints” of greenhouse gases in climate data is statistically invalid. He has also provided a slightly less technical summary of the densely mathematical original at Judith Curry’s blog. But wait, you say, he’s not a climate scientist. No, he teaches econometrics and it turns out the methods in question are important statistical tests regularly used in econometrics. Moreover, he published his paper in the same journal that published the original statement of the method. So the usual dodges won’t work. His claim is that 20 years’ worth of studies connecting climate changes to greenhouse gases have been done in such a flawed way that no one knows if they are right or wrong.
What do the scientists whose method he trashes have to say? They are sort of silent, but not entirely. McKitrick reports that he has been sending his criticisms to them for 2 years, and has interacted with them through the journal review process and by email. The two scientists are Myles Allen of Oxford and Simon Tett of the University of Edinburgh. He offered to delay publication of his critique until they had a chance to write a reply, but neither took him up. Although Tett eventually responded, encouraging him to go ahead and publish the work.
The paper McKitrick criticizes is nicknamed AT99, after the authors and the year of publication. Amidst a detailed analysis of mathematical errors in the AT99 method, which as he points out has been the cornerstone of IPCC reports over the years, he makes the following accusations:
[AT99] made claims about the properties of their estimator that, at best, were never proven and in general are not true. Their [method] ... violates an important sufficient condition for unbiasedness.
[There] is no assurance that any applications of the AT99 method have yielded unbiased coefficient estimates... claims about the unbiasedness of optimal fingerprinting regression coefficients based on the AT99 method are conjectural.
Thus it is not possible to assign meaning to [the results of using AT99].
Inferences based at least in part on the AT99 methodology have driven some of the most consequential policy decisions in modern times... Confidence in the results of the AT99 methodology rests on their claim that it ... yields unbiased and efficient coefficients, and that as long as the [AT99 test score passes] the variance estimator is valid. These claims are untrue.
There is obviously a lot to unravel in this story. But we predict the alarmists’ affection for peer-reviewed science will evaporate long before they have to accept this paper at face value, nor will they be able to resist the temptation to dismiss a paper about statistical theory by someone trained in statistical theory because he is not a climatologist, even though they are happy to endorse papers about statistical theory by climatologists untrained in statistical theory. Stay tuned.
I'm not at all qualified to opine on the technicalities of attribution methodologies but I think it's fair to assume that legitimate debates could be had about the soundness or innumerable variables and methodologies baked into the climate models.
This is not a sausage anyone wants to see being made as long as it tastes satisfyingly alarmist when it comes out.