Positive predictive value (PPV) for simulated scenarios varying in power, bias and prevalence of true effects (i.e. above the minimum of interest).

Scenario descriptions, parameter combinations and the Analytical column are adapted directly from Ioannidis [1]. The remaining columns contain the PPVs for simulations of 5,000 published findings using each of the different effect size distributions in our model. Parameters for the distributions are set so that the prevalence of effects above the minimum of interest approximates the specified odds (see S1 Table for model parameters and achieved prevalences for each scenario). Missing values occur in some scenarios when the desired prevalence cannot be achieved without changing the minimum effect of interest, thus complicating direct comparisons with other scenarios. * For Two Normals, one cannot obtain a prevalence of 66.7% without changing the minimum effect of interest or the SD of the wider normal, as a normal N(0, 1) has only about 62% of its mass beyond the minimum of interest on both tails. ** For Two Peaks, one cannot obtain a prevalence of 0.1% without changing the minimum effect of interest, as a normal distribution N(0, 0.2) has over 1% of its mass beyond the minimum of interest on both tails.