Just a Sample

April 3, 2009
Posted by Jay Livingston

Obama has nominated a sociologist to head the census bureau, Robert Groves (summa at Dartmouth, degrees in statistics and sociology from Michigan, where he is now a professor). According to the New York Times, the choice of Groves, “instantly made Republicans nervous.”

The problem is scientific sampling, also known as the possibility that we might actually get an accurate count of the kinds of people the census usually undercounts.
Republicans expressed alarm because of one of Mr. Groves’s specialties, statistical sampling — roughly speaking, the process of extrapolating from the numbers of people actually counted to arrive at estimates of those uncounted and, presumably, arriving at a realistic total.
The Republicans favor a census that tries to count each person individually, an obvious impossibility with a systematic bias in favor of people who are in places that make them easy to count.

I remember an anecdote from a book on sampling – I wish I could remember the author and title – about a social scientist who had done some research on soldiers and was presenting his finding to a general in the Pentagon. The officer questioned the idea of sampling. How could you know about the hundreds of thousands of soldiers in the military by talking to a couple of thousand. How many should we talk to, asked the sociologist. “You gotta do ’em all,” said the general.

“General,” said the sociologist. “When you go to the doctor, he takes a little tube of blood to find out how much cholesterol and other things are in your blood. That’s a sample. Do you tell him that if he really wants to know the true amount, he has to take it all?”

No comments: