Weighted Bootstrap Inference on Censored Regression Model

Regression model with censored data is studied. Based on synthetic data, weighted bootstrap is used to approximate the sample distribution of regression coefficient. For general choices of random weights, the validity of such a distribution approximation and the weak consistency of weighted bootstrap estimate to regression coefficient are established. A weighted bootstrap confidence interval for regression coefficient is then constructed. Simulation studies were conducted to compare the weighted bootstrap interval with the interval based on normal approximation and showed that the former is superior to the latter in terms of coverage probability.