Felix Singleton Thorn
CAMARADES, University of Melbourne
Co-Authors: Paul Dudgeon, Fiona Fidler
Statistical power and effect sizes in psychology are decreasing over time
This poster presents a meta-analysis of 46 studies assessing the statistical power of psychology research at Cohen’s effect size benchmarks, and an analysis of over 130,000 effect size estimates from over 9,000 articles published in 5 APA journals from 1985 to 2013. The first study shows that the average statistical power of psychology is extremely low for ‘small’ effects, .23 (95% CIs [.17, .29]), somewhat low for ‘medium’ effects, .62 (95% CI [.54, .70]), and only acceptably high for ‘large’ effects, .80 (95% CI [.68, .92]). It also shows that these values have changed little if at all over time, with an estimated yearly change of -0.000 (95% CI [-0.003, 0.003]), 0.001 (95% [-0.002, 0.004]), and -0.001 (95% [-.002, 0.001]) at the small, medium, and large benchmarks. However, the results of the second study show that effect sizes reported in published psychology research are becoming smaller over time, suggesting that the average statistical power of psychology research is decreasing.