Inference for correlated effect sizes using multiple univariate meta-analyses

摘要: Multivariate meta-analysis, which involves jointly analyzing multiple and correlated outcomes from separate studies, has received a great deal of attention. One reason to prefer the multivariate approach is its ability account for dependence between estimates same study. However, nearly all existing methods meta-analytic data require knowledge within-study correlations, are usually unavailable in practice. We propose simple non-iterative method that can be used analysis meta-analysis datasets, no convergence problems, does not use correlations. Our uses standard univariate marginal effects but also provides valid joint inference parameters. The proposed directly handle missing under completely at random assumption. Simulation studies show unbiased estimates, well-estimated errors, confidence intervals with good coverage probability. Furthermore, found maintain high relative efficiency compared conventional meta-analyses where correlations known. illustrate through two real functions estimated interest.