Statistical Inference for Estimators in a Semiparametric EV Model with Linear Process Errors and Missing Responses

Abstract

This paper concentrates on the properties of estimators in a semiparametric EV model, particularly considering the effects of missing data and linear process errors according to the actual situation. The missing data are processed by three different methods: the direct deletion method, imputation (interpolation fill) method, and regression surrogate method. Also, the corresponding estimators of the slope parameter β and the nonparameter variable g ⋅ are obtained. All the estimators are asymptotically normal, and the consistency rates for which can achieve o n − 1 / 6 log n . Besides, the performance of the estimators is investigated by one sample experiment.