Abstract
The missing-rank approach and missing-failure time approach of estimating unknown parameter of the Cox proportional hazards model for the interval censored data are compared in this article. In missing-rank approach, it does not require specification of the baseline hazard function and use a Gibbs sampling scheme to generate rankings for estimating parameters. In missing-failure time approach, the parameters of the baseline function are estimated simultaneously with the regression parameters, and the estimating equation is solved using Monte Carlo techniques. Further using the asymptotic result of Satten et al. (1998), we study the local confounding effect of interval censored data which may cause the estimator bias. Finally, we show some simulation results of comparison and their discussion in three types of data set.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . 33 Estimation of β . . . . . . . . . . . . . . . . . . . . 4 3.1 Missing-rank approach . . . . . . . . . . . . . . . 4 3.2 Missing-failure approach . . . . . . . . . . . . . . 94 The local confounding effect . . . . . . . . . . . . . . 125 Simulation . . . . . . . . . . . . . . . . . . . . . . . 196 Discussion . . . . . . . . . . . . . . . . . . . . . . . 277 Appendix A . . . . . . . . . . . . . . . . . . . . . . . 298 Reference . . . . . . . . . . . . . . . . . . . . . . . 33