混合治癒模型之文獻回顧

Abstract

臨床醫學研究的資料常可見某些患者並不會經歷感興趣的事件。即使有充裕的觀察時間，但“免疫者”依舊存在的存活分析稱為治癒模式。混和模型是最常見的建模方法，母體被區分為“可致病”和“免疫”的兩個群體，並分別對於“免疫比例”與 “可致病者之發病時 間”做進一步假設。本論文中回顧此領域部份重要文獻，探討母數與半母數的架構下之推論方法，包含最大概似估計法與動差法。我們提供系統化的角度，透析不同方法背後的建構原則，希望對未來延伸的研究有所助益。

Cure models are suitable for analyzing survival data when some people never experience the event of interest despite of long-term follow-up. The most popular modeling approach is the so-called mixture model in which the population is divided into a susceptible group and a group of cure. In the thesis, we review important literature on the cure mixture model. Both parametric and semi-parametric inference methods are considered. In particular, the likelihood approach and methods based on some moment relationship are examined under different model assumptions. We aim to provide a systematic way of studying and comparing these methods.