Multiple comparisons, interaction effects, and statistical inference: lessons from chronic kidney disease progression among blacks.

Abstract

A post hoc analysis of a randomized trial comparing progression of chronic kidney disease among blacks and non-blacks provides an opportunity to explore statistical inference. Multiple comparisons without penalizing the P-value can lead to false positive results; this is illustrated using simulation. Tests of statistical interaction are then applied and interpreted to understand effect modification (or lack thereof) in the context of this trial.