UC Merced

When researchers conduct a meta-analysis, the sampling variances of individual study effects are important for estimating the meta-analytic model. However, for some effect size measures such as correlations and standardized mean differences, the variance is dependent on the true effect size in the population, which is not normally known. This study uses a new “dynamic method” for mixed-effects meta-analysis for correlations and standardized mean differences that takes this dependence into account. Comparing results using the dynamic method with other commonly used meta-analysis methods in the frequentist and bayesian framework shows that the dynamic approach has improved estimation, particularly with correlation effect-size. Another finding of note is that incorporating Hedges bias correction (1981) standardized mean differences is shown to increase bias in conventional meta-analytic results.