December 22, 2010 § 2 Comments
I’m starting this blog because I have little to show for too much time spend arguing about statistics on the Internet. The thing is, I think that arguing about statistics on the Internet can have a positive effect.
Frank Sulloway and Richie Zweigenhaft published a paper on the relationship between birth order and risk-taking among Major League Baseball players. In the past, when two (or more) brothers have made it to the majors, the younger brother is a bit more likely to have the higher rate of attempted steals in the family. Sulloway and Zweigenhaft were worried about the confounding effect of ability, so they controlled for birth order, comparing older brothers called up first to younger brothers called up first, and older brothers called up second to younger brothers called up second. Now the difference is huge (odds ratio 10.58).
Phil Birnbaum and his commentariat thought this was weird at best, and to me it seemed like a Type M error. To Sulloway and Zweigenhaft’s credit, they responded to the concerns raised. To me, though, the main concern—the weird control—isn’t convincingly dealt with. My thoughts are strewn across several comments on Phil’s blog. I’d like to edit them before reproducing them here. Maybe next week.