Fig. 3From: The normality assumption on between-study random effects was questionable in a considerable number of Cochrane meta-analysesProportions of significant non-normality in different subgroups for meta-analyses when using A ORs, B RRs, and C RDs. On the horizontal axis, the first line presents restrictions on sample sizes, and the second line presents restrictions on event rates. The corresponding numbers of meta-analyses satisfying the restrictions are in parentheses. The vertical bars represent 95% confidence intervals of the proportionsBack to article page