針對半競爭風險與截切資料估計累積發生函數

Abstract

本篇論文針對半競爭風險與截切資料探討累積發生函數的無母數估計方法。發現文獻中已有Peng & Fine (2006) 討論過這個問題，並以“分解代入法”提出估計量。我們先分析競爭風險與半競爭風險資料所提供訊息之差異，並探討截切限制對估計造成的影響，最後以“加權修正偏誤”的方法提出新的無母數估計量。藉著電腦模擬實驗我們比較兩種估計量的表現，並探討造成優劣之可能原因。

In this thesis, we consider nonparametric estimation of the cumulative incidence function based on truncated semi-competing risk data. Peng & Fine (2006) apply the idea of decomposition to construct an estimator. Under a similar framework, we propose an alternative weighting approach. The two estimators are compared via simulations. The results show that our proposed estimator performs better which is possibly due to the fact that we make slightly stronger assumptions. We also investigate the difference between the information given by competing risks data and that given by semi-competing risks data.