A simple multiply robust estimator for missing response problem

A multiply robust estimator for a missing response problem is recently proposed that is more robust than doubly robust estimators proposed in the literature. Its formulation is based on empirical likelihood, which solves an implicit Lagrangian equation and often encounters computational problems such as multiple roots or nonconvergence. An alternative multiply robust estimator is proposed, which is computed by least squares and can be implemented easily in practice. We show that this multiply robust estimator is locally semiparametric efficient. Copyright (C) 2013 John Wiley & Sons, Ltd.