I've had a closer look at the "bombshell" paper that Patrick Michaels described like this:
After four years of one of the most rigorous peer reviews ever, Canadian Ross McKitrick and another of us (Michaels) published a paper searching for "economic" signals in the temperature record. ... The research showed that somewhere around one-half of the warming in the U.N. surface record was explained by economic factors, which can be changes in land use, quality of instrumentation, or upkeep of records.
There seems to be some problems with their work. To understand them you need to understand the two different ways of measuring angles.
|This angle is one degree.||This angle is one radian.|
Can you spot the difference?
If you do calculations and get degrees and radians mixed up, you get the wrong answer. Which is what McKitrick did. His analysis included a variable
cosablat, which was supposed to be the cosine of absolute latitude. Trouble is, the software he used expects angles to be measured in radians, his data has latitude in degrees, and he didn't convert from degrees to radians. Consequently, every single number he calculates is wrong. I corrected the error and reran his regressions. The sizes of the "economic" signals were greatly reduced. They no longer "explain" half of the surface warming trend. Removing the effects of the economic variables now just reduces the warming trend for his sample from 0.27 degrees/decade to 0.18 degrees/decade, which is very close to the warming trend for the whole globe.
Even this overstates his results---McKitrick did not calculate statistical significance correctly---his analysis incorrectly assumes that each observation comes from a different country. His "economic signals" may not even be statistically significant.
Somehow these errors were not detected during the "four years of one of the most rigorous peer reviews ever". Nor did peer review by Climate Research detect the problems. Unfortunately, this is not the first time that the peer review process at Climate Research has failed. Last year, several of the editors resigned after another defective paper slipped through peer review. Oddly enough, that paper also attempted to cast doubt on anthropogenic global warming.
If you're new here: In previous postings on Ross McKitrick I have shown how he messed up an analysis of the number of weather stations, showed he knew almost nothing about climate, flunked basic thermodynamics, couldn't handle missing values correctly and invented his own temperature scale.
Update: John Quiggin confirms my findings.
This article (scroll to page 13) by Clare Goodess has more on the problems with the review process at Climate Research that led to the resignation of half of the editors. Chris de Freitas, the same editor that published the previous improperly reviewed papers, also published the McKitrick and Michaels effort.
Update 28/8: A poster on the Internet Infidels forums demonstrates how easy it is to verify that McKitrick got it wrong.
Jaw-dropping. If I may ask, how did you come to notice this?
So I'm guessing that "four years of one of the most rigorous peer reviews ever" means something more akin to "four years of angry threats and bribes"...
Tim, have you considered submitting a paper pointing this out to CR?
It's not surprising they'd have a problem with cosines, given that they've never fully accepted round earth theory.
An interesting question is how they came to use cosines in the first place. My guess is that they tried absolute latitude and got results they didn't like. So they fiddled with the specification and found one that gave them what they wanted. Tim, if it's easy, could you do a run with abslat and see what happens.
Good 'un Kuas.
Woe to the person who incurs Tim's wrath, that's what I say. Well done, sir. Keep up the good work.
I expected this blog to be informative but never thought it would be quite so funny (in a really frightening way).
Lesson: Never take a knife half-baked theory to a gun data fight :-)
If it weren't so sad, I'd be laughing. Good work, Dr. Lambert.
alkali, I had a feeling that their results would go away if a control for latitude was added. When I went to add it, I found that it was already there, but with an odd way of calculating the cos (by using sines). I wondered why, so I looked up the SHAZAM manual and found that it doesn't have cos, just sin, but also that the angle had to be in radians. I'd already looked at his input data and knew latitude was in degrees.
John, I did a run with abslat and the results are very similar to the corrected version with cos abslat. It does seem suspicious that they would use cos abs lat. After all, that is the same as cos lat. Why do cos abs lat unless they had started with abs lat?
There is more dodgy stuff in their data. E.g. I didn't know that Antarctica had a GDP.
Upon further reflection, Tim, I think Ken Miles has a great point.
Why have this information languish in the blogosphere? It would do a lot of good in the non-electronic-osphere as well. A nice headline in the Australian, for instance (right, Ken?) would do a world of good.
The reaction by the vested interests to your attempts at publishing this work likely would make a few good essays...
Tim, Great work! Here's another question. When you decided you wanted to check McKitrick's analysis, how did you locate the actual code and datasets he used? I haven't seen the "third" paper yet, so I don't know if there was a link to it there. If I wanted to do something like this on my own in the future, where would I start? Thanks again! - Scott
I left the original query at Prof Q about the TCS bombshell claims with a genuine desit\re to expose it to criticism and think Tim has done a great job exposing the problems. In the past Ive done similar things with the FU Nature paper with others who have diffrent viws.
Can I also ask whether Tim has any comments about the recent posting in New Scientist about sunspot cycles and cloud cover, and whether the IPCC habe tested sun cycles or controlled for the in their various computer simulations. There seems over varous time scales evidence that sun activity cycles may be important drivers of climate change, or at least correlate significantly
An interesting piece of work: congratulations. Some comments:
You should inform both the authors and the publishers of what you have found. Its a sufficiently gross error to be easily checkable, so they should be able to respond quickly. It would be unkind not to tell them.
As an aside, McK seems to be pushing the "four years of one of the most rigorous peer reviews ever" as a virtue of the paper. Of course it isn't. You don't explicitly point this out, I think, but 4 years in review mean that it started off as a truely awful paper and it has struggled hard to come up to any kind of standard since. They are trying to make a virtue out of a flaw. The editorial/review correspondence for this paper would make very interesting reading if available!
I guess without a master mathematician/statistician like Essex at his side, McKitrick is totally lost at sea ;)
Four years in review means the editor and referees should get co-authorship. And in fact,that's not a bad idea: one would know all the participants in the screwup.
That is good counsel, Mr Connolley. I appreciate your comments elsewhere, and now here.
'4 years' is aimed at the rubes demographic, I think...
For the record my attention on this matter is formally noted. Given my work duties, don't expect an early reply, but make sure your science is sound and replicable.
"For the record my attention on this matter is formally noted. Given my work duties, don't expect an early reply"
I think Louis will reach retiring age before he comes up with a reply on this one.
John, You might find a reply here, and in Henry, and as your reply is easily classed as an ad-hominem, anything else I may use is permissable and accessible.
Louis has announced that this issue is serious, everyone. Use spellchecker and sit up straight at your keyboards...*snork*...ahem...when replying.
Given my work duties, don't expect an early reply, but make sure your science is sound and replicable.
FYI, it took me just a few minutes to verify that Tim Lambert's claims are "sound and replicable". A thorough review of Tim's work and a reply from you regarding this matter shouldn't take more than 10 or 15 minutes of your valuable time.
And as for mixing up degrees with radians, I can tell you first-hand that calculating with degrees when radians are called for *really* messes up the results.
If my understanding of the correction is correct, the claimed one-half economic signal of the warming trend is really one-third.
If so, who here is claiming that is insignificant?
0.07 is less than one-third of 0.27, TJ. I think you may have looked at 0.07/0.20, in which case your understanding of the correction is incorrect.
Actually, that 0.20 should have been 0.18 so it was a third. McKitrick didn't calculate the standard errors correctly so this economic signal may not be statistically significant and may be really caused by other geographic causes that have not been controlled for.
More importantly, McKitrick is trying to use his paper to argue that most of the warming trend in the surface record is not real -- he is trying to show not just that there is an economic signal, but that it dominates, and this certainly is not true.
It's not that McKitrick is the first to make such an error, but it is probable that he is the first to publish. Getting the wrong unit as input to a subroutine is common. Checking is a duty, and this is one of the first things you look at in student work, if, for nothing else, to inculcate in them the proper sense of paranoia.
They can use you guys at the EPA. I think they are tired of creating their own rules, and just pulling numbers out of the sky. One example is that second hand smoke more harmful the exhaust. lol
How do you expect peer review to find errors like this? I expect this sort of thing happens a fair amount. Good work on finding the error, although I think you are very silly with how you express it...and that making an error does not mean that one needs to hang up one's cleats.
Did the authors admit the error and issue a correction, with acknowledgement to you? If not, they should have. If they have, you ought to note it, to finish the story off.