Thursday, December 2, 2010

The Republican War on Statistics

Ah, sample size. We all know that sample size can be a real problem. Too small of a sample, and you get all sorts of spurious results. This is especially true if your population is huge but your sample is small.

But what would you say about a study with a population of about 400,000, that actually sampled a full 28% of the population? Oh my god, this is a statistician's wet dream. A sample that large and that comprehensive, assuming there are no major selection biases, you can basically get the precise answer for the entire population. It would take an astounding coincidence for that to be misrepresentative sample.

And yet... and yet... In the Senate hearings on DADT that are going on right now, at least three Rethuglican senators (McCain, Inhofe, and Brown... oh joy, tools all around) have expressed concern that the Senate committee report on the possible repeal of DADT "only" sampled 28% of the 400,000 or so active military personnel. Inhofe even said that 28% "isn't very much."

Wow. I mean, even for these lying sacks of shit, that's just... just... I'm speechless.

This would be like a baseball player, who hit a home run every third time he swung at a pitch. "Hmmm, I dunno, one in three isn't very much. I don't think that guy's very good at hitting home runs..." Oy.

1 comment:

  1. Heh, I'm looking forward to using this in my research methods class next term. Very few of my students will ever go on to become researchers, but I hope the class helps inoculate them against tendencies to believe that kind of bullshit.