Fig. 8
From: Synthesising results of meta-analyses to inform policy: a comparison of fast-track methods
Proportion of wrong conclusions obtained with COMA, MAMA, REMA, and SOMA, with and without publication bias. Three types of publication bias are considered, bias 1 (only individual studies showing significant effects are published), bias 2 (only individual studies with significantly negative effects are published), and bias 3 (only individual studies with significantly positive effects are published). The computations were performed assuming a true mean effect size equal to − 0.29 (loss of − 25%, left), − 0.69 (loss of − 50%, middle), or zero (no effect, right)