Statistics is Beautiful?
From reader Yeah, Yeah comes a link to a Wired article which assures us we should “Learn the Language of Data.” It’s not a pretty language, but it can be remunerative.
Only problem is, the writer makes the common mistake about weather and climate. Heat waves are evidence of global warming, but snowstorms are not. Still, it’s nice to see people care about people like me.
Reader Chuck Lampert asks about the increase in the number of temperature “records”, as discussed briefly in this article.
An increase in the number of records can be caused by many things, not just increasing temperatures. More records (as time goes on), better observations, more or different locations, etc. are all likely to reveal more records.
Gist: just seeing a new “record” is not direct evidence that temperatures have warmed.
Reader Al Perrella reminds us of an old paper by William Jeffreys and Jim Berger: “Sharpening Ockham’s razor on a Bayesian strop.” They show how the ad hoc principle of parsimony made famous by William of O can be seen in a Bayesian light.
Both Jeffreys and Berger are subjective Bayesians and believe that probability is purely a matter of opinion. Berger in particular focuses on decision theory, where that kind of behavior is natural. However, as I’ve pointed out many times, subjectivists and objectivists (like myself) rarely disagree on any actual analysis.
Fox News Climate Story
Gene Koprowski, a columnist at FoxNews.com (well, that’s what he called himself), wrote last week and asked me to comment on this: “Addressing the growing worries in the public, and declining credibility, due to concerns about how the U.N. gathers and spreads climate data, the international body [IPCC] is now assembling a “more diverse” group of authors for its newest report. But, rather than ideological diversity, the team features geographic diversity, meaning, critics tell FoxNews.com, more of the same in elite opinion on climate change.”
I replied: If “embracing geographic diversity” means merely selecting IPCC members to fill proportional quotas from country or region, then obviously this is silly as a means of improving climate science.
Don’t forget that the *majority* of IPCC members are not climate scientists or meteorologists, etc.; they are economists, social scientists, and so forth. Thus, this “diversity” move may just reflect a desire to boost research of regional climate change effects, not climate change itself.
The problem is that it’s the *climate science* which is under contention, not the economics, etc. Instead of bringing in geographically diverse people, they should bring in accredited scientists who have made legitimate criticisms of the models. We need a more diverse group of physicists, climatologists, and meteorologists. And we need more people (I blush to say) people like me, statisticians whose specialty is quantifying uncertainty.
Everybody on the IPCC assumes that the models are good. Why? No tests of predictive performance have been made. Compare the accuracy of your local weather forecast—which is usually at least reasonable—with that of the predictions made by climate models (and environmentalists). How many of them have been right? We need to turn climatology into a more honest science with better measured uncertainty in its predictions.
Beware of Cow’s Milk
Longtime reader Ken Fischer points us to yet another example of bureaucracy run amok. The EPA has “classifie[d] milk as oil“. Excuse me?
“[T]he Environmental Protection Agency (EPA) is classifying milk as oil because it contains a percentage of animal fat, which is a non-petroleum oil.” And we all know how bad oil spills can be.
Yes, dear reader, the United States of America has devised a new law where we now legally must cry over spilt milk.
Random Walk Climates
Professors Guang Wu and Shaomin Yan sent me a link to their site, Dream Sci Tech, where they play with some models which express the view that climate change is best modeled as a random walk.
I haven’t had a chance to look at the site in any depth, but a glance at the other papers published by the pair shows that they know their way around a computer (particularly for models of amino-acids and mutations).
Cheating with SATs
Reader Scott Bury shows us how easy it is to cheat with statistics, particularly when the victim wants to be cheated. People so love stories of bias that they never look too deeply into them.
So it isn’t any surprise that the media ran with a report that the “SAT is biased against minorities”, without bothering to check that the claim had any merit (it does not).
Reader Randy Brich points us to Matt Ridley’s newest: The Rational Optimist: How Prosperity Evolves. Anybody read this yet?