Tools for Those Who Summarize the Evidence Base
I have multiple effect sizes (i.e., correlations between the two continuous variables of interest) per study. Two questions:
1. When constructing my funnel plot of effect size (r) plotted against the inverse variance weight, I am keeping all of the effect sizes in the graph for analyses of funnel plot asymmetry. Is there a consensus on if this is appropriate or if I should randomly select one per study?
2. I'm worried my nested effect sizes violate the regression assumptions used in the Egger test and Peters test of funnel plot asymmetry. If anyone could point me towards best practice in how one might look statistically at funnel plot asymmetry with effect sizes nested within studies, I would be most appreciative!