Psychologists Change the Way They Analyze Their Data ?

Daryl J. Bem. Cornell University Jessica Utts and Wesley O. Johnson University of California, Irvine We agree with Wagenmakers, Wetzels, Borsboom, & van der Maas (2011) that there are advantages to analyzing data with Bayesian statistical procedures, but we argue that they have incorrectly characterized several features of Bem’s (2011) psi experiments and have selected an unrealistic Bayesian prior distribution for their analysis, leading them to seriously underestimate the experimental support in favor of the psi hypothesis. We provide an extended Bayesian analysis that displays the effects of different prior distributions on the Bayes factors and conclude that the evidence strongly favors the psi hypothesis over the null. More generally, we believe that psychology would be well served by training future generations of psychologists in the skills necessary to understand Bayesian analyses well enough to perform them on their own data.