Simulations for a Q statistic with constant weights to assess heterogeneity in meta-analysis of mean difference

Abstract

A variety of problems in random-effects meta-analysis arise from the conventional $Q$ statistic, which uses estimated inverse-variance (IV) weights. In previous work on standardized mean difference and log-odds-ratio, we found superior performance with an estimator of the overall effect whose weights use only group-level sample sizes. The $Q$ statistic with those weights has the form proposed by DerSimonian and Kacker. The distribution of this $Q$ and the $Q$ with IV weights must generally be approximated. We investigate approximations for those distributions, as a basis for testing and estimating the between-study variance ($\tau^2$). Some approximations require the variance and third moment of $Q$, which we derive. We describe the design and results of a simulation study, with mean difference as the effect measure, which provides a framework for assessing accuracy of the approximations, level and power of the tests, and bias in estimating $\tau^2$. Use of $Q$ with sample-size-based weights and its exact distribution (available for mean difference and evaluated by Farebrother's algorithm) provides precise levels even for very small and unbalanced sample sizes. The corresponding estimator of $\tau^2$ is almost unbiased for 10 or more small studies. Under these circumstances this performance compares favorably with the extremely liberal behavior of the standard tests of heterogeneity and the largely biased estimators based on inverse-variance weights.