How not to defend your statistics

[wshaffer]

So, the most recent episode of the More or Less podcast tackles The Spirit Level, a book that I reviewed fairly favorably not too long ago. The book purports to show that societies with greater income equality do better on a whole bunch of measures. More or Less presenter Tim Harford begins by noting that there's a suspicious correlation between someone's politics and how critical they are of the statistical methods used in the book. And I have my suspicions about where Tim Harford's politics are. (He writes for the Financial Times.)

However, in her interview with Harford, Kate Pickett, one of the books coauthors, comes across as a bit of a twit less intellectually rigorous than one would like. I mean, misremembering details of particular data points is understandable, but responding to "Why didn't you do a multivariable analysis?" with "Because we didn't think the other variables were important" makes you sound like you're assuming the hypothesis you're trying to prove. Sigh.

Remember people: If you're going to make sweeping statistical claims that nicely confirm my political prejudices, please be rigorous!