This artical investigates estimation of the coe?cients in measurement er-ror model under MAR with missing data . In the first part of the dissertationWe obtain the H-T weighted estimator by using kernel smoother and logisticmodel to estimate the selection probabilities which are unknown. In the secondpart of the dissertation we discuss the properties of the weighted estimatingequation in a sense of imputation , furthermore we propose using nonparamet-ric model and semiparametric model to estimate the conditional expectationin this estimating equation. At last, we present large sample theory for thenew estimator and conduct simulation studies. |