Testing homogeneity of many within-subject variances using small samples

Abstract

SUMMARY The problem of testing homogeneity of many within-subject variances using small samples is addressed. The asymptotic null distribution of minus twice the likelihood ratio statistic, as the number of subjects tends to infinity, is derived. By inverting the characteristic function of this statistic numerically, the performance of the corresponding critical values may be assessed, and compared with competing approximations. A real data example and a theoretical comparison demonstrate the accuracy of the critical values derived.