Weighted local linear approach to censored nonparametric regression

This article proposes a unified and easily implemented nonparametric regression method for estimating the regression function for censored data under both iid and time series contexts. The basic idea, of the method is to use a constructed weighted local likelihood by,combining the benefits of the local polynomial fitting. The estimation procedure is implemented and a bandwidth selection criterion, based on the generalized cross-validation criterion, is proposed. Further, the finite sample operating characteristics of the proposed method is examined through simulations, and its usefulness is also illustrated on two real examples. Finally, the consistency and the asymptotic normality of the proposed estimator are established, which provide an insight into the large sample behavior of the proposed estimator. In particular, the explicit, expression for the asymptotic variance of the resulting estimator is given and its consistent, estimate is provided.