Hacker News Reader

Why is it "hokey" to do register studies where you explore correlations and test various confounders?

I'd say it is an extremely valuable , indispensable even, item in the scientific toolbelt.

Because asking a group of people how often they sunbathe is a horrible way to measure sun exposure. Dishonest answers, imperfect memory, differing definitions of sunbathing, and the idea itself that sunbathing and tanning is a proxy for overall sun exposure when they likely don't even make up a significant fraction of a person's UV exposure. Bad, bad, bad.

Because Correlation Is Not Causation™, and if they're doing correlation against self-reported data that involves people estimating things it's even less reliable. It still has value, yes, but that value is in finding things for more rigorous methods to examine, not in trying to draw conclusions from such shaky materials.

Reply on news.ycombinator.com