http://www.erikdrysdale.com/winners_curse/
Erik Drysdale
bioeconometRician
About Recent Posts ShinyApps CV
A winner’s curse adjustment for a single test statistic
“One consequence of using frequentist statistical tools to conduct scientific inference is that effect sizes will be biased when conditioned on statistical signifiance, even if the test itself is unbiased…. The power-bias relationship helps to explain the Proteus phenomenon whereby follow-up studies tend to have a smaller effect size.”