Wednesday, October 04, 2006

The limits of case-control studies: breast feeding and intelligence

Everyone who's ever written about breast feeding and intelligence presumably knew that IQ is largely determined by parental genetics and/or early environment and presumably tried to control for that. It appears they didn't control well enough:
Breast-feeding has no impact on intelligence

... The researchers found that although breast-fed children scored higher on IQ tests this was because their mothers tended be more intelligent, better educated and provided a more stimulating environment at home....
What's interesting here is not the apparently lack of effect of breast-feeding on IQ. That was always highly suspect and breast-feeding appears to have other advantages anyway (though I suspect those are overstated for cultural-political reasons). Nor is the inevitable confusion between correlation and causation particularly interesting. That's so commonplace it's boring.

No, the interesting part is that despite enormous statistical sophistication, we still have a lot of trouble drawing reliable conclusions from case-control studies. I wonder if a case-control study result is really any more robust than extrapolating from experimental studies on animal populations. Maybe we need to downgrade case-control and upgrade experiments involving non-human animals...

No comments: