SEMIPARAMETRIC ESTIMATION OF MODELS FOR MEANS AND COVARIANCES IN THE PRESENCE OF MISSING DATA

In this article we describe a class of inverse-probability-of-censoring-weighted estimating equations for jointly estimating the parameters of models for the conditional mean and covariance of a vector of responses given a set of regressors in the presence of monotone missing outcome data. Our methods are valid when the data are missing at random in the sense of Rubin (1976) and do not require a parametric model for the joint distribution of the data. However, they do require a model for the non-responsive probabilities, We show that the solution to the optimal estimating equation in our class has asymptotic variance equal to the semiparametric variance bound, Because the optimal estimating equation depends on unknown population parameters, we propose an adaptive locally efficient estimator whose asymptotic variance can achieve the semiparametric variance bound.