Analyses of effect size inflation.

Analyses to detect evidence of effect size inflation using different ways to handle favorable/unfavorable outcomes and meta-analyses with most of the sample size preceding the highly cited study. In “Publication order”, we did not invert the ratio if a replication had most or all of its sample size published before the highly cited study. In “effect coining”, we used unfavorable outcomes for all effect measures, thus avoiding the need to reverse ratios for favorable ones. The bottom row combines both approaches. P-values refer to a one-sample Student’s t test against a theoretical mean of 0 for the log-transformed inflation ratios for each study pair. Values for means and 95% CIs were calculated for these log-transformed ratios and exponentiated back to the original scale. Inflation point estimates were manually replicated with the results of the R code output (https://osf.io/9hmx4) and this analysis can be found at https://osf.io/qnfgh.