FIGURE 3.

Results of the Bayesian bias correction (BBC) reanalysis of the religious priming studies. In the (left), we display the prior (gray) and posterior (black) probabilities of eight possible scenarios that result from the combination of four possible bias processes and two states of nature (to wit, that there is a real effect [+] vs. that there is no effect [–]). The greatest positive change, and the largest resulting posterior probability, is seen in model 3+. This model includes a biasing process whereby non-significant results are published at a smaller, but nonzero rate. Additionally, it includes a true effect. The high posterior likelihood on this model implies that a true effect exists. In the (right), we display the prior and posterior distributions of this effect size. The gray dashed line and gray bar indicate the prior distribution of the effect size, with a 50% probability of it being exactly 0 and a 50% probability of it being somewhere in the domain of the normal distribution. In the (black) posterior distribution, the point mass at 0 has vanished as the mass of the posterior has accumulated around 0.3. As a result, the Bayes factor favoring the null hypothesis is practically 0, indicating strong evidence for the alternative hypothesis that there is an effect of religious priming.