MULTIVARIATE LOGISTIC-MODELS FOR INCOMPLETE BINARY RESPONSES

In this article we describe a likelihood-based regression model approp riate for analyzing incomplete multivariate binary responses. We focus on ''marginal models''; that is, models where the marginal mean or ex pectation of the binary response is related to a set of covariates. Th e association between the binary responses is modeled in terms of cond itional log odds ratios. When the nonresponse mechanism is ignorable, it is not necessary to specify a nonresponse model, and valid inferenc es can be obtained provided that the likelihood for the responses has been correctly specified. But when the nonresponse mechanism is nonign orable, valid inferences can only be obtained by incorporating a model for nonresponse. An unresolved issue with nonignorable models concern s the identifiability of the parameters. So far, no general and practi cally useful necessary and sufficient conditions for identifiability a re available. Here we suggest some simple procedures for examining the identifiability status of nonignorable models when the response varia ble is discrete. Finally, we present results for an analysis of multip le informant data from the New Haven Child Survey and the Eastern Conn ecticut Child Survey.