Testing for Dichotomous Moderators in Meta-Analysis

Publisher Website
Google Scholar
Abstract
The authors conducted Monte Carlo simulations that compared Type I error rates and the statistical power of 3 tests in detecting the effects of a dichotomous moderator variable in meta-analysis: (a) the Q B test proposed by L. V. Hedges and I. Olkin (1985), (b) the z RR test proposed by R. Rosenthal and D. B. Rubin (1982), and (c) the z HS test proposed by J. E. Hunter and F. L. Schmidt (1990). Those procedures differ in selected effect size index and in the weighting scheme for each effect size. Number of primary studies, average sample size, and distribution of the parametric effect sizes were the manipulated conditions. The z HS test showed the highest statistical power, although at the cost of an inadequate inflated Type I error rate in meta-analyses with a low number of studies. The Q B and z RR tests adequately adjusted the Type I error rates, and their statistical power was similar. Criteria for selecting among the tests are discussed.