Can I get an AMEN!!! Great article by Zed Shaw on how programmers, or anyone who knows how to spell “standard deviation”, is a statistics guru. I can’t stand when I have to listen some numb-nut ramble on about some analysis he did, and I know he’s gotten not only the method, but the underlying assumptions all wrong. Statistics is the only science where a moron can take a semester of, or even a weeks training, and think they know it all.
I love it when I can tear these people down.
Great read…enjoy (warning: some foul language will be found in the attached article):
“Programmers Need To Learn Statistics Or I Will Kill Them All” by Zed Shaw
Great article by William M. Briggs that sheds new light on how to view the recent article regarding the “independent” analysis performed by several statisticians on global temperatures, which allgegedely “confirms” that global cooling is not occuring. He is absolutely correct when he states that:
“Since we [statisticians] are free to choose from an infinite bag, all of our models are suspect and should not be trusted until they have proven their worth by skillfully predicting data that has not yet been seen. None of the models in the AP study have done so. Even stronger, since they said temperatures were higher when they were in fact lower, they must predict higher temperatures in the coming years, a forecast which few are making.”
The famous statistician, Stu Hunter, says that getting at the underlying trend in a time series is like trying to identify a criminal by looking at “smudged fingerprints”. Time series modeling is an iterative process and in the end you can only hope that you’ve gotten at the underlying trend. It is more art than science in some respects.
The ultimatel goal of time series analysis is to adequately forecast future observations (i.e. predict). Was the data these statisticians reviewed partitioned into two data sets, the model buildt on one, and then used to predict the other set? Did they all agree on the final model, independent of one another? How do the models differ? As in any publicized analysis, can their approach to the data be vetted by the statistical community? I’m not attempting to plug holes in the story, but the use of statistics has to be careful in these instances given the stakes, and should be subject to the same peer review rigor as a journal article.
William M. Briggs Article
Great question/point by William M. Connolley in Stoat:
“Are we really supposed to believe that the statisticians who received this data didn’t recognise it? You have to be fairly out of touch not to know what the global temperature record looks like.”
As both a scientist and statistican, I still think we need to consistently keep George Box’s famous quote in the back of our minds as we look analyze any dataset and reach conclusions, “All models are wrong, but some are useful”. This is a great quote from the father of Time Series analysis, which I presume is the method used to analyze this data.
Statisticians are pretty smart folks. My guess is that at least one may have recognized the trend. But here’s where I get real geeky. Assuming that each statistician was completely independent of the other and that there was a 25% chance that each one independently would have recognized the data (TOTAL guess on this 25% probability), there is a total .0039% chance that all four would have recognized it. Adjusting the individual probability lower would obviously make the total probablity lower and vice versa.
So why the easy math above…I don’t think all four would have recognized the data for what it was and their conclusions are fair.
With that being said, they should release the same exact dataset so all of us statisticians can all take a look at it!!! This is what you would do with any data used to back up any scientific conclusions. Open the same exact data up for inspection, analysis and discussion.
AP IMPACT: Statisticians reject global cooling?
