Comparing Three Approaches to Handling Dependency: A Case Study of a Meta-Analysis of Writing Self-Efficacy and Writing Proficiency

Abstract

This study compared three approaches (i.e., averaging within-study effect sizes, three-level meta-analysis, and robust variance estimation) to handle dependent correlational effect sizes in conducting a meta-analysis. Data were from a meta-analytic study examining the relationship between writing self-efficacy and writing proficiency. To examine the differences in the performance of the three approaches, seven conditions were created by the number of studies and the number of effect sizes per study. While all three approaches produced similar results in the average effect size and standard error, the averaging approach had much smaller variance estimates. The patterns were basically consistent across different conditions. This study informs meta-analysts of appropriate procedures in handling the dependent effect sizes.