本篇論文中主要是利用最大概似法對長期追蹤資料和事件時間建立聯合模型來處理與時間相依的共變數,其模型中包含了生物指標的長期追蹤模型以及事件時間的疾病風險模型。在長期追蹤資料的部分使用的是線性混合效應模型。在存活資訊的部分,由於有母數的方法對於數值之集中趨勢、分散性及分佈所提供的資訊有較好的解釋能力,因此在此考慮使用有母數聯合模型來進行資料的分析,並且考慮存活模型為較具彈性的擴充風險模型,擴充風險模型包含AFT模型及Cox模型為此二模型之廣義模型。在Cox模型中,只有最近一次的共變數會影響實驗對象的風險,而AFT模型則是考慮到整個共變數歷史對於風險的影響。因此,在擴充風險模型中,共變數透過風險因子以及基準風險函數分別包含了瞬間的影響以及累積的影響,所以此擴充風險模型也會有助於描述存活分析中複雜的生物議題,而在本篇論文中將會分別給定Weibull分配及Log-logistic分配之基準風險函數來進行聯合模型對於時間相依共變數的模擬分析以及對地中海果蠅資料進行實例分析,並且使用EM演算法、Newton-Raphson法及蒙地卡羅法等數值方法來進行參數的估計。;In this paper, we used the maximum likelihood approach to jointly model the longitudinal and event time processes. This involves selecting a longitudinal model for the biomarkers and a disease risk model for the event time data. For the longitudinal component, we use linear mixed effects model. For survival component, we use parametric hazard model because parametric hazard model provides better interpretation for concentrate trend, dispersion and distribution. And we consider the survival model is the more flexible model, extended hazard model, contains Cox model and AFT model. In the Cox model, only the covariate at the most recent time has an impact on the subject’s specific risk while the accelerated failure time model allows the entire covariate history to influence the risk. In the extended hazard model, the covariates have both instantaneous impact and cumulative impact on the risk through the risk factor and the baseline hazard, respectively. Therefore, we will let baseline hazard function is Weibull distribution or Log-logistic distribution to simulate data and analyze the Mediterranean fruit fly data and estimate parameter by EM algorithm, Newton-Raphson and Monte Carlo method.