An Investigation of the Raudenbush (1988) Test for Studying Variance Heterogeneity.

Abstract

Abstract The meta-analytic method proposed by Raudenbush (1988) for studying variance heterogeneity was investigated. This method uses the difference in two weighted log-standard deviations (e.g., treatment vs. control group) as an effect size and provides a test of the homogeneity of effect sizes. The relationship between study characteristics and effect sizes can also be investigated. Although appealing, the performance of this procedure for realistic data sets characterized by nonnormal data and small study sample sizes is unknown. The results of a Monte Carlo study indicated that the Type I error rate of the test was sensitive to even modestly platykurtic score distributions and to the ratio of study sample size to the number of studies, even if the data were normally distributed. On the whole, this test cannot be recommended to meta-analysts.