ABSTRACT
Givenbinarydatafromatwo-stageclustersample,weoutlineandapply amethodofBayesianpredictiveinferenceforafinitepopulationproportion,P.Ourprobabilisticspecificationshouldbeusefulformanysurveys ofthistypeandyieldssimpleanalyticalexpressionsforthemomentsof thepriorandposteriordistributions.Tofacilitateroutineanalysisofsurveydata,weinvestigateapproximationsfortheposteriordistributionof P.WecontrastourinferenceswiththoseusingempiricalBayestechniques anddesign-basedmethods.Weapplyourmethodologytoasurveydesignedtoinvestigatethequalityofcarethatcancerpatientsreceive.By appropriatespecificationofthepriordistributionweshowthatthereis agreementbetweentheobserveddataandproposedmodel.
