競爭風險下串行事件發生率之迴歸分析

Abstract

在此論文中，我們探討帶有競爭風險的串行事件資料，特別感興趣的是在第二段中發生特定風險的發生率。串行事件資料的間隔變數會發生相依設限的問題，也進而影響到發生率的估計。針對傳統的競爭風險資料，已有文獻針對於累積發生函數以及發生率建立迴歸模型，我們將現有方法推廣到串行事件資料。利用加權方式解決相依設限所造成的偏誤，並且進行模擬實驗檢驗在有限樣本下的性質。我們並將所提出的方法分析一筆廔管栓塞的復發資料。

In this thesis, we consider serial events data in the presence of competing risks. Specifically we focus on estimating the incidence rate of a particular type in the second stage in which induced dependent censoring occurs. There exist some regression models for the cumulative incidence function (CIF) under the classical setting of competing risks. Here we extend the method proposed by Chang and Wang (2009) to the second-stage estimation for serial events data. Specifically we apply a weighting approach to handle the problem of induced dependent censoring and conduct simulation analysis to examine the finite-sample performances of the proposed method. We also apply the proposed method to analyze a real dataset of shunt thrombosis recurrences.