ABSTRACT
The idea of synthesizing simple effect sizes such as standardized mean differences, correlation coefficients, and odds ratios has been extended to a set of multivariate approaches. These can be used to combine effect sizes in complex models. When the effects are correlations, multivariate techniques allow for mediation models, path analytic models, and structural equation models. This chapter first reviews asymptotic distribution theory for the elements of the correlation matrix and functions thereof. It then presents univariate and multivariate methods for combining correlation matrices and using them to estimate path models. It also describes some newer techniques that enable reviewers to extract and synthesize partial effect measures directly from regression analyses in primary studies. An example is used to illustrate the procedures.
