Output details
11 - Computer Science and Informatics
University of East Anglia
On the moments of Cochran's Q statistic under the null hypothesis, with application to the meta-analysis of risk difference
<27>We generalise the approach in Output 2 to a more realistic setting. In Output 2, the effect and weight for an individual study depended on a single parameter, but here they may depend additionally on a nuisance parameter. Derived expansions, without any normality assumptions, for the first two moments of Q will have wide applicability in testing for homogeneity in meta-analysis. A particular case of our theory is the popular Welch (1951) test. As another important example, we present a homogeneity test for the risk differences which is substantially more accurate than the currently used test.