Fig. 6: Null hypothesis significance tests as transformations of variables.
From: Bayesian p-curve mixture models as a tool to dissociate effect size and effect prevalence
a In a previous study (n = 54), we estimated the prevalence of above-chance performers on a discrimination task by modeling the distribution of participants’ accuracies as a mixture between a Binomial distribution for participants for whom \({H}_{0}\) was true (at-chance accuracy) and a Beta-Binomial distribution for participants in which \({H}_{1}\) was true (above-chance accuracy). b We could have instead converted the accuracies to p-values using a binomial test and modeled the distribution of p-values as a mixture between two p-curves. c Both models result in almost identical posterior distributions for the population prevalence and d identical per-participant posterior probabilities of \({H}_{1}\).
