Wednesday, April 01, 2009

Vinyl flooring, autism and statistics

So now there are thoughts that vinyl flooring causes autism. I really like this aspect of the study: "The researchers relied on questionnaires and did not measure any chemicals in the homes, which limits the reliability of the findings because they do not know for certain that the children were exposed to phthalates." Despite the authors' own caveats (calling it "far from conclusive..."), this report, as with most environmental health reports is an endless source of frustration to me. So many unexamined variables, so few comparisons to other studies, so little prediction and follow-up studies.

Overriding all of this is the standard significance test of P = 0.05. This means that even in the best statistical designs, there is a 5% chance that you are wrong in your conclusions. Not because you did something wrong or because someone has a political agenda opposing the results or ... but because statistical results are just that - statistical, and 1 out of 20 times you're going to end up rolling snake eyes.

Think about that next time you pick up a medical journal with 20 different studies in it.

No comments: