Regression Analysis Of Dependent Interval-censored Failure Time Data With The Additive Hazards Model

Interval-censored data is an important type of censored data which frequently occur in survival data analysis. For this type of data,most of existing research assumes the censoring is independent or noninformative, that is, the censoring time and the event failure time are independent. However, this assumption does not some-times hold in real applications. At this case, ignoring the dependence between these two variables will make the estimation biased even wrong. There have been some research which employed Cox's model to analyze the informative interval-censored data. Different from Cox's model, the additive hazards model is another kind of important model in survival analysis, and in this model, the covariate effect takes the form of additivity and describes the absolute influence of the covariate on the hazard function. So far, there have been relative less researches about additive haz-ards model and most of which assume the independent censoring mechanism. In this thesis, we discuss the regression analysis of dependent interval-censored data under the additive hazards model. A common latent variable was used to describe the dependence between the censoring time and the event failure time. The maximum likelihood estimator was derived and the asymptotic properties of the proposed esti-mator were also given. Finally some extensive numerical simulations were conducted to evaluate the performance of the proposed approach. The simulation results show that the proposed procedure is reasonable and effective.