Abstract

Despite years of criticism, null hypothesis significance testing (NHST) continues to be psychology's most widely-used model for the purposes of statistical inference. Since Fisher (1925), psychologists have increasingly adopted the model in ruling out chance from their conclusions. The purpose of the present investigation was to address two questions: (1) what are the problems associated with NHST and are they due primarily to inherent difficulties of the model, or due to misunderstanding and misuse, and (2) given the problems that accompany the use of NHST, can any other model successfully replace it, and thereby overcome these problems? It was found that although many problems can be attributed to user misuse, there are nonetheless serious theoretical and methodological flaws inherent in the model itself that warrant the search for a substitute model of inference. Upon considering alternatives, it was found that confidence intervals, Loftus' plot-plus-error-bar procedure (PPE) and Serlin & Lapsley's “good-enough principle” could feasibly serve as replacements to NHST. The Bayesian model of statistical inference, despite its alleged problem of prior probabilities, is recommended as the “best” alternative to NHST. Furthermore, the fact that posterior probabilities converge is used as support for why even frequentists should approve of the Bayesian approach.