Dr. Adam Hafdahl

Department of Mathematics, Washington University

Title: Meta-Analysis for Functions of Correlations

Abstract: Conventional meta-analysis of correlations entails combining and comparing independent studies' estimates of one bivariate association. When two or more such associations are synthesized, meta-analysts often treat each one in isolation, which cannot address patterns defined by more than one association. In this talk I will introduce strategies suited for certain multiple-association questions. After describing inference for functions of one study's correlation matrix (e.g., partial or multiple correlations, regression coefficients) and multivariate meta-analysis for correlation matrices, I will show how to analyze such functions based on a correlation matrix estimated from several studies.