ESTIMATION OF A COMMON ODDS RATIO UNDER BINARY CLUSTER SAMPLING

We present results from a simulation study for the estimation of a com mon odds ratio in multiple 2 x 2 tables when the data are correlated w ithin clusters. We model the correlation of the data by the beta-binom ial distribution. Through a simulation study, we compare the Mantel-Ha enszel estimator with Rao and Scott's estimator in terms of their bias es, observed variances, relative efficiencies of their variances and 9 5 per cent coverage proportions. We limit the simulation study to the case where there are the same number of subjects in each cluster and t he same number of observations in each raw of each stratum. When rho = 0, we recommend use of the Mantel-Haenszel estimator gamma(MH) with a n unadjusted variance and Rao and Scott's estimator gamma(RS)(P) with a pooled design effect. In general, when rho > 0, we recommend the Man tel-Haenszel estimator gamma(MH) with an adjusted variance and Rao and Scott's estimator gamma(RS)(P) with a pooled design effect.