REGRESSION-MODELS FOR LONGITUDINAL BINARY RESPONSES WITH INFORMATIVE DROP-OUTS

This paper reviews both likelihood-based and non-likelihood (generaliz ed estimating equations) regression models for longitudinal binary res ponses when there are drop-outs. Throughout, it is assumed that the re gression parameters for the marginal expectations of the binary respon ses are of primary scientific interest. The association or time depend ence between the responses is largely regarded as a nuisance character istic of the data. The performance of the methods is compared, in term s of asymptotic bias, under misspecification of the association betwee n the responses and the missing data mechanism or drop-out process.