Within-study covariance estimators for network meta-analysis with contrast-based approach

Abstract

The contrast-based approach is one of the primary approaches in network meta-analysis. For statistical modeling in network meta-analysis and meta-regression models, within-study covariance estimates are needed to adequately address the correlations among the multivariate outcomes. In this computational note, we present the formulas of covariance estimators for standard effect measures used in modern meta-analysis practice: risk difference, risk ratio, odds ratio, mean difference, and standardized mean difference (Cohen’s d and Hedge’s g).