Empirical Likelihood Inference of Linear Transformation Models with Right Censored Data

摘要

This article develops empirical inference method for linear transformation models with randomly right censored data. An empirical likelihood ratio statistic is constructed through a synthetic data approach and is shown to have a limiting weighted chi-square distribution. For inference convenience, an adjusted empirical likelihood ratio statistic is proposed on base of the former one, which is shown to have a limiting standard central chi-square distribution. Confidence regions of regression parameters are then constructed. A simulation study is carried out to investigate the performance of the empirical likelihood method and the adjusted empirical likelihood method proposed in this article compared with the traditional normal approximation method. It results out that the two empirical likelihood methods have more accurate confidence regions and better coverage probabilities than normal approximation method.