3 ms·
Generally you can control for this statistically; the study yields information of the sort you need, but you're correct that it's not totally causal. But there
by tedks 12y ago
Generally you can control for this statistically; the study yields information of the sort you need, but you're correct that it's not totally causal.
But there aren't experiments for most things we'd like to know anyway. Experiments are a nice thing to have, but they're not necessary to gather information and update your beliefs rationally.
- grayclhn 12y agoI would say sometimes you can control for this statistically. When the selection problem is exactly along the dimension you want to measure, (i.e., measure the impulse-control of children of parents with poor impulse-control, but only the part caused by angry yelling) statistical correction is less plausible.