Bayesian Meta-Analytic SEM of Dependent Correlation Matrices

• In Event: Advances in Meta-Analytic and Systematic Review Approaches

Abstract

We present a Bayesian approach for meta-analytic SEM (MASEM) of correlation matrices. The approach is based on the transformation of sample correlation matrices to an unconstrained vector, a vector that is likely multivariate normal. This approach allows for modeling fixed- and random-effect models, and can also handle dependent correlation matrices, such as clusters of matrices resulting from the same authors within a single MASEM application. The approach is implemented in the bayesianmasem R package, and we demonstrate its application to a real world example. The approach is currently unable to handle missing data, but we identify a path towards resolving this challenge.

Authors

• Nduka Boika, University of North Texas

• James O. Uanhoro, University of North Texas