In analyzing partially observed categorical case-control data, William
son and Haber (1994, Biometrics 50, 194-203) assumed ignorable nonresp
onse. We reanalyzed the data allowing for nonignorable nonresponse whi
le explicitly incorporating case-control sampling.