Running the Numbers

One advantage and drawback of working in a building full of statisticians is never being the smartest person in the room. I wrote in amusement yesterday about the St. Louis Fed positing that belief in Hell correlated with lower corruption and higher per capita income ("Get back to work."). This morning, I received an email from a few cubes over:

Oh, BTW- on that hell/corruption thing, the p-value was 0.047. I saw that last night & found their raw data. Of course, they failed to control for all kinds of covariates that could explain what's really going on... case in point, instead of "fear of hell" substitute "porn" and you'll get about the same correlation.

[Here is] the original article, data and graphs. A quick read shows that they didn't look at any alternative hypotheses or really bother with covariates. I don't have time at the moment, but you should always be suspicious when they go to ranks when their distribution to start with is pretty good.

Sweet.

Update (8:10 pm): Kevin Drum and Daniel Drezner weighed in on the subject. Two word summaries: "pretty weak" (DD) and "simplistic crap" (KD).