Bootstrapping Cox’s Regression Model with Time-dependent Covariates

Utilizing the counting process technology and the von Mises method, we discuss the asymptotic properties of bootstrapping the Cox’s regression model for censored survival data with time-dependent covariates which have a proportional effect on the hazard function of the life-time distribution of an individual. Under some regular assumptions, we show that bootstrapping the model is feasible: both the maximum partial likelihood estimators for the regression coefficients and the nonparametric estimator for the baseline cumulate hazard rate function is consistent.