A Study of a New Effect Size for Variance Heterogeneity in Meta-Analysis

• In Event: 33.058 - Effect Sizes, Sample Sizes, and Power

Abstract

Variance heterogeneity has traditionally been treated as a nuisance in two-group studies interested in examining means, despite the fact that patterns of heterogeneity can provide important information about the impact of a treatment. Similar opportunities exist in meta-analysis but an effect size capturing variance heterogeneity in two-group designs with desirable statistical properties has not appeared in the literature. The difference in two coefficients of variation is introduced as an effect size along with tests of homogeneity. A Monte Carlo study of the statistical properties of the resulting statistic and tests for realistic data conditions is underway.

Author

• Michael R. Harwell, University of Minnesota