72
reads

Jin-Jian Hsieh

2010-10-22
12:45:00 - 14:45:00

405 , Mathematics Research Center Building (ori. New Math. Bldg.)

Semi-competing risks data are commonly encountered in medical studies, which are consist of two types of events, namely terminal and non-terminal. For semi-competing risks data, since the non-terminal event may be censored by the terminal event dependently, we can't make inference on the non-terminal event without extra assumption. Therefore, we assume that the dependence structure on non-terminal event and terminal event follows a copula model and let the marginal regression models of non-terminal event and terminal event both follow time-varying effect models. We use a likelihood approach to estimate regression coefficient of non-terminal event. We also prove the large sample properties of the proposed estimator. Via the simulation studies, it shows the performance of our proposed method is well. We also apply our method into a real data which is AIDS Clinical Trial Group (ACTG 320) for illustration.