Experiment-Wise Type I Error Control: A Focus on 2 × 2 Designs

Abstract

Factorial designs are common in psychology research. But they are nearly always used without control of the experiment-wise Type I error rate (EWER), perhaps because of a lack of awareness about viable procedures for that purpose and perhaps also because of a lack of appreciation for the problem of Type I error inflation. In this article, key concepts relating to Type I error inflation are discussed, with emphasis on the 2 × 2 factorial design. Simulations are used to evaluate various approaches in that context. I show that conventional approaches often do not control the EWER. Alternative approaches are recommended that reliably control the EWER and are simple to implement.