Cryptic multiple hypotheses testing in linear models: overestimated effect sizes and the winner's curse.

Cryptic multiple hypotheses testing in linear models: overestimated effect sizes and the winner's curse.