2011-08-012024-05-13https://scholars.lib.ntu.edu.tw/handle/123456789/646147摘要：在許多臨床醫學與流行病學的研究中，研究個體於追蹤期間可能會發生多元有序非致命事件或甚至死亡，在此情況下，終止事件死亡會停止非致命事件過程之發生。對此一序列非致命事件過程是否對發生下一個新的事件和終止事件具有預測效力，經常是研究者的研究目的之一，因此估計非致命事件之間與終止事件之間的相關性可作為判斷非致命事件過程是否具有預測能力的首要統計方法。所以本研究的研究目的為針對具終止事件之多元有序非致命事件資料將推廣Kendall’s tau 與交叉比值等相關測度，以為建構多元非致命事件之間以及其和終止事件之間與事件順序及時間有關的各種相關性測度，並由此導出得到其無母數估計方法，本研究亦將探討所提出估計方法的統計性質，並將以模擬說明及比較本研究所提出各種估計方法在有限樣本時的表現，最後以實際資料說明本研究所提出各種估計方法的用處。<br> Abstract: In many clinical and epidemiological studies, subjects may experience a sequence ofmultiple nonfatal events and may die during the follow-up period. In the situation, theterminal event death censors the observation of the nonfatal events process. It is often ofinterest to identify the predictive ability of the nonfatal events on the risk of a new nonfatalevent and the terminal event. Assessment of the associations between the nonfatal events andthe terminal event can be used for evaluating the prognostic utility of the nonfatal events.The episode-specific and time-varying association measures between the nonfatal events andthe terminal event will be introduced by modifying the Kendall’s tau and the cross ratio. Thegoal of this study is to develop nonparametric estimation for the episode-specific andtime-varying association measures for the sequential nonfatal events data with a terminalevent. The asymptotic properties of the proposed methods will be developed. Thecorresponding finite-sample performance will be evaluated in a simulation study. Thepractical usefulness of the proposed methods will be illustrated in an analysis of a data set.