Fig. 7.
The magnitude of each meta-analysis' estimated effect size declines after correcting for publication bias. Nine of 20 meta-analyses of lnRR, 17 of 36 meta-analyses of SMD and 14 of 31 meta-analyses of Zr had corrected directions of slope after adjusting for publication bias. The remaining 11 in lnRR, 19 in SMD, and 17 in Zr had the wrong direction of slope, presumably because of a high degree of heterogeneity that could not be controlled for. Original = uncorrected meta-analytic estimate effect sizes (i.e. βo[overall] in Equation 1). Bias-corrected = meta-analytic estimate effect size corrected for the presence of two forms of publication bias, small-study and decline effects (i.e. β0[bias − corrected] in Equation 3)