本研究的目的在於針對存活資料中含有的長期追蹤(Longitudinal) 性質的共變數(生物指標)，並且伴隨著測量誤差的情形下去計算時間相依一致性指標(Time-dependent concordance)。我們進一步比較的三種方法可以依據基於模型(Model-based)與非基於模型分成兩類，基於模型中包含了聯合模型(Joint model) 補值法與鄰近點補值法(Nearest Neighbor Estimate)，而目前存在的文獻中的非基於模型則是利用IPCW(Inverse of the probability of censoring weighted)來進行加權。本研究會比較此三種方法所估計的時間相依一致性指標，在各種不同測量誤差、測量值缺失率、設限率及樣本數下的影響，最後以實際愛滋病的資料做分析，展示三種方法下的結果。;The purpose of this study is to calculate the Time-dependent concordance for the Longitudinal covariance (biological index) contained in the survival data, and to calculate the time-dependent C-index with the measurement error. The three methods we further compared can be divided into two categories based on the Model-based and non-model-based. The model-based method includes the Joint model compensation method and the Nearest Neighbor Estimate. The non-model-based methods in the existing literature use IPCW (Inverse of the probability of censoring weighted) for weighting. This study will compare the time-dependent C-index estimated by these three methods, and the impact of various measurement errors, missing measurement rates, censoring rates, and sample sizes. Finally, the actual AIDS data will be used to analyze the three methods.
