Here is top notch excellent reporting in the WaPo on a recent study that correlates coffee drinking with living longer.

Here is why the article is top notch:

“Careful, though — this doesn’t prove that coffee makes people live longer, only that the two seem related. Like most studies on diet and health, this one was based strictly on observing people’s habits and resulting health. So it can’t prove cause and effect.”

Indeed, the study cannot prove anything. The best it can do is correlate. Kudos to the reporter for making this so explicit (it’s only a few paragraphs down from the lead and it’s pretty unambiguous).

There are probably a dozen (or more) things that you might be able to tease out of the data set that correlate in “interesting” ways. Of course, they wouldn’t be that interesting because without causality you’ve got…well, nothing.

Now, we just need to get people to stop thinking that doing correlation studies are useful…and to get people to stop thinking that correlations are useful…

-JD Cross

Advertisements