Principles of Model Specification in ANOVA Designs

Abstract ANOVA—the workhorse of experimental psychology—seems well understood in that behavioral sciences have agreed-upon contrasts and reporting conventions. Yet, we argue this consensus hides considerable flaws common ANOVA procedures, these become especially salient the within-subject mixed-model cases. The main thesis is are model specification. specifications underlying use deficient from a substantive perspective, is, they do not match reality experiments. problem, particular, rely on coincidental rather than robust statements about reality. We provide avoid making arguments based coincidences, note Bayes factor comparisons among already convenient BayesFactor package. Finally, specification necessarily critically reflects concerns, and, consequently, ultimately responsibility researchers. Source code for project at github/PerceptionAndCognitionLab/stat_aov2 .