don't discriminate much, take every idea, the odd idea that gots strong reaction (in a good, strange, nonobvious way), probably is on to something.

same idea with anova, the variable that causes a lot of variance is probably significant.