p-Curve and Selection Methods as Meta-Analytic Supplements for Biologists: A Demonstration of Effect Size Estimation in Studies of Human Fluctuating Asymmetry

Fluctuating asymmetry is hypothesized to predict developmental instability (DI) and fitness outcomes. While published studies largely support this prediction, publication bias remains an issue. Biologists have increasingly turned to meta-analysis to estimate true support for an effect. Van Dongen and Gangestad (VD&G) performed a meta-analysis on studies of fluctuating asymmetry (FA) and fitness-related qualities in humans. They found an average robust effect size, but estimates varied widely. Recently, psychologists have identified limitations in traditional meta-analyses and popular companion adjustments, and have advocated for alternative meta-analytic techniques. P-curve estimates true mean effects using significant published effects; it also detects the presence of p-hacking (where researchers exploit researcher “degrees of freedom”), not just publication bias. Alternative selection methods also provide a means to estimate average effect size correcting for publication bias, but may better account for heterogeneity in effect sizes and publication decisions than p-curve. We provide a demonstration by performing p-curve and selection method analyses on the set of effects from VD&G. We estimate an overall effect size range (r = 0.08–0.15) comparable to VD&G, but with notable differences between domains and techniques. Results from alternative estimation methods can provide corroborating evidence for, as well as insights beyond, traditional meta-analytic estimates.