Another project idea: Meta-analytic methods for correlational data

Several different approaches have been proposed for meta-analysis of correlation coefficients. One of the major differences between approaches is the choice of scale: whether effect sizes should be analyzed on the Pearson-r scale or first transformed to the Fisher-z scale. This project will study methods for modeling correlation coefficients on the r scale in the presence of between-study effect heterogeneity. Specific topics include:

• refined methods for variance estimation;
• hierarchical modeling to capture differences between distinct operationalizations of the same construct; and
• application to a large correlational meta-analysis.

This project would be appropriate for a Quantitative Methods graduate student with interests in meta-analysis and hierarchical models.