Fig. 4
Variation of the most appropriate p cut-off value for 3 effect sizes (d), 3 prior odds of H1 relative to H0, and different sample sizes in each group (n). Only results that provide an α ≤ 0.05 and a study power ≥ 0.8 are presented. Note that the x-axis has a logarithmic scale. The y-axis is −log(p); therefore, as we go upper on the axis, the p value decreases. The horizontal dot-dashed gray line corresponds to the conventional p cut-off value of 0.05. Note that for some designs, there is a minimum sample size to comply with the constraints imposed on the α and study power. For example, it is not possible to discover a difference with an effect size (d) of 0.2 with a sample size of 500 per group, if a prior odds of H1 relative to H0 of 0.1 is assumed (the solid red line)