Saturday, March 03, 2007

Grumbling Over Statistical Stumbling

From Glenn Reynolds, here's a fascinating Economist story about statistical bumbling in medical studies. It's provocactively subtitled "Why so much medical research is rot."

He's talking about those studies where they hunt through medical databases, searching for correlations between conditions. Apparently the researchers have trouble with rules of statistical inference when testing multiple hypotheses.

Hey, I would have trouble too. Probability and statistics can be counter-intuitive. It helps to have a real guru handy.

Could this explain why medical recommendations about what to eat keep changing so much?

When searching for natural laws
Beware of correlation without cause.

No comments: