在醫學研究中有幾種常見的研究設計, 包含病例對照研究、世代研究及臨床實驗。而本論文選擇的設計是結合病例對照研究以及世代研究的病例世代研究設計, 並且把驗證集的概念放入研究設計中, 以達到將成本降低的效果。在本文中我們考慮在病例世代研究設計下, 運用Bickel et al. (1993) 提出的偏誤抽樣方法, 結合Chen (2002) 提出的驗證集抽樣方法。透過參數化的存活迴歸模型並結合上述的統計方法, 得到參數的估計方程式,進而能夠探討解釋變數與存活時間之關聯。最後透過資料模擬及參數估計的偏誤、變異數和覆蓋概率評估所提出的統計方法論。
In many medical studies, sampling schemes are implemented to save time and cost for large scale of epidemiological studies, including case-control study, cohort study, and validation sampling. In this thesis, we consider that the sampling design which is the combination of case-control study and cohort study for time-to-event data. In order to take the biased sampling scheme into consideration, the probability of being selected into sample for each observation is derived under the Weibull distribution. Furthermore, part of subjects is able to collect the accurate covariates without measurement error such that the estimates of regression coefficients have lower variance. So the estimating equations of parameter are developed through the maximum likelihood approach for data of this type. In simulation studies, the performance of the proposed method is examined by comparing the bias, standard error and coverage probability. We find that validation sampling does improve the statistical properties of parameters.