Effect size distributions used in the simulations.

In the Dichotomous model (A), the density curve is replaced by a histogram, as effects sizes are strictly 0 or 1. The Two Peaks (B), Overlapping Peaks (C) and Two Normals (D) models are built as a mixture of two normal distributions with different means and/or SDs (see Methods for details). The weight parameter that regulates the mixture is 0.5 in the examples shown, leading to equal weights for both curves–in simulations, however, it varies according to the prevalence of true effects. For the Single Normal model (E), the weight parameter equals 1 and the standard deviation is used to control the prevalence. Dashed lines indicate the minimum effect size of interest (held fixed at 0.5 for all simulations): effects are considered true if their absolute size is larger than this threshold.