LAD Estimation for Multiple Regression Models with Doubly Censored Data

Abstract

Least absolute deviations (LAD) estimator for multiple regression models with doubly censored data is proposed and the asymptotic normality of the estimator is established. The proposed method doesn't require the censoring vectors to be identically distributed and the estimator can be applied to models with either fixed or random design. Moreover, as a generalization of the LAD estimator for models with complete observed data, the proposed estimator is robust to heteroscedasticity of the errors.