An effect size for variance heterogeneity in meta-analysis

Abstract

The difference in two coefficients of variation is proposed as an effect size to capture variance heterogeneity for two groups across a sample of studies, and five tests of homogeneity of this effect size were studied. Capturing within-study differences in variability across a sample of studies should enhance understanding of treatment effects estimated by traditional mean-based effect sizes by providing information about the extent to which subjects in a treatment group responded similarly across studies relative to a control group. Monte Carlo results suggested the Miller-Feltz test can be used to test homogeneity of within-study variability.

Keywords:

• Variance heterogeneity
• Meta-analysis
• Monte Carlo study