| Summary: | 碩士 === 國立臺灣大學 === 流行病學研究所 === 96 === In the development of the medical science, the multi-state data consisting of the course of disease progression are frequently encountered in longitudinal studies. The semicompeting risks data is a type of multi-state data where an intermediate event may be censored by a terminal event. When the terminal event is subject to left truncation, the naive regression analysis for the intermediate event based on the competing risks data in the presence of left truncation, only use part of data and the information of the observed intermediate event may be excluded by artificial truncation which may lead to large efficiency loss. For estimating the regression parameters in the relative risks model of the cause-specific hazard function for the intermediate event, estimation methods using all intermediate event information are developed under the situations of the independent and dependent terminal events respectively. Simulation studies are conducted to compare the performance of the proposed and naive estimators. Finally, we also apply those methods to analyze a colon cancer data set.
|
|---|
