Usefulness of the p_rep statistic as an alternative to null-hypotheses significance testing using p values

Comparisons of analytical results based on p_rep with p

Abstract

In a recent article, Killeen (2005) proposed the statistic p_rep, the probability of replicating an effect, as an alternative to traditional null-hypotheses significant tests (NHST). In this article, two experiments were conducted and their analytical results based on p_rep and traditional p values were compared: non-significant results based on p values were reinterpreted as meaningful results in light of power analysis, calculating the effect size and p_rep value. The tendency of p value analyses not to reveal non-significant results (i.e. the file drawer problem) and the improvement of decision-making methods are discussed.