本研究主要在使用統計軟體R，進行時間相依解釋變數對於存活時間(survival time)影響的評估，本文採用 Cox 比例風險模型。釋變數外，本文考慮在加入了隨機截距項(random intercept)於時間相依解釋變數上的情況下生成存活資料，以達到每位病患都各自擁有自己的解釋變數。透過模擬分析，本文探討了R 裡的函數'coxph' (survival package) 在多種狀況之下於估計迴歸係數時的表現。
此外，在配適Cox模型的解釋變數時，除了使用一般處理時間相依解釋變數的方法，亦考慮忽略時間相依解釋變數，單取起始量測值為時間獨立解釋變數進行模型配適，探討是否對於估計有任何影響。最後以成大醫院腎臟科一項慢性腎病之研究進行實際配模操作，並且以該方法偵測可能增加風險的因素。

This study focuses on assessing statistical software R for survival data in the presence of time-dependent covariates. The common Cox model is considered to illustrate the association between the hazard function and the potential factors. The time-dependent covariates, for instance some biomarkers, of each patient are recorded at each clinical visit. The survival time is subject to usual right-censorship. In the simulation studies, many different types of time-dependent covariates, including linear process, nonlinear process, and process with individual effect, are assumed in the generating procedure of survival times.
In addition to the usual fitting process in R with the function 'coxph' with time-dependent covariates, models with initial values of covariates taken at the onset time are also considered in the simulations. By comparing the simulation results, we can explore the impact of including the whole observed time-dependent covariates in model fitting. This method is then applied on a Chronic Kidney Disease (CKD) study in the National Cheng Kung University Hospital to detect the possible risk factors.

1.Austina, P.C. (2012). Generating survival times to simulate Cox proportional hazards models with time-varying covariates. Statist. Med. 31,3946–3958
2.Cox, D.R. (1972). Regression models and lifetables. J. Royal Stat. Soc. Ser. B 34,187–220
3.Cox, D.R. (1975). Partial Likelihood. Biometrika,Vol.62,No. 2,269-279
4.Johansen, S. (1983). An Extension of Cox’s Regression Model. International Statistical Review 51,258-262