Here we show that W&H's conclusion is false, i.e., that publishing everything is more effective than selective publishing. We use ‘effective’ to encompass statistical considerations (i.e., precision of the estimate of meta-analyses), and cost-benefit and time considerations. First, we demonstrate that publishing everything is more effective, using exactly the same simulation results and scenario as used by W&H [1]. Then we show the superiority of publishing everything over selective publishing in a scenario where the population effect is zero and the researcher's goal is to detect small (Cohen's d = .2) or larger effect sizes (d>.2).
