Application of Convariance Process to Tests for Censored Paired Data

Application of Convariance Process to Tests for Censored Paired Data

the covariance process of two martingales provides a useful tool to capture the dependence structure for paired censored data. in this paper it is applied to modify the variances of weighted logrank tests in order to take account of dependence between paired subjects. In the process of modification a 'variance correction term' is introduced. Some variance estimators based on separate samples are considered together. Performance of the estimators are compared through simulation studies. Several independence tests for bivariate sruvival date are also proposed which are naturally reduced from the weighted logrank tests accomodating dependence structure. Simulation studies are carried out to compare the independence tests. Both the weighted logrank tests and the independence tests are illustrated by an example.