Estimation methods are proposed for fitting logistic regression in which outcome and covariate variables are missing separately or simultaneously. One of the two proposed estimators is an extension of the validation likelihood estimator of Breslow and Cain (1988). The other is a joint...