Sieve maximum likelihood estimation for the proportional hazards model under informative censoring

Failure time data often occur in many areas such as clinical trails, economics and medical follow-up studies, and a great deal of literature has been developed for their analysis when the censoring is noninformative. A number of methods have also been developed for the situation where the censoring may be informative. However, most of the existing procedures for the latter case apply only to limited situations or may not be stable or robust. In this paper, we present a copula model approach for regression analysis of right-censored failure time data in the presence of informative censoring. In the method, the copula model is used to describe the dependence between the failure time of interest and censoring time and for estimation, a sieve maximum likelihood estimation procedure is developed. In addition, the asymptotic properties of the proposed estimators are established and the simulation study indicates that the proposed method seems to work well in practice. An illustrative example is also provided. (C) 2017 Elsevier B.V. All rights reserved.