在本論文中，主要透過先前研究所發展出的一種預測精準度指標concordance來作為衡量疾病預測能力的標準，它是藉由研究中提出應用在固定共變數下的時間相依敏感度與特異度所導出並使用Cox模型描述共變數與存活時間的關係。然而使用Cox模型需要資料符合比例風險假設，若不符合假設我們可以利用其他模型像是加速失敗（AFT）模型來替代。接著，本研究更進一步將其推廣到具有長期追蹤的生物指標上。我們藉由後續的模擬章節來評估本篇論文所推廣的程式其結果表現，以及根據三筆實際的資料來展示出推廣後的成果。;In this thesis, we used a concordance index as a measure of disease prediction ability which was derived from time-dependent sensitivity and specificity with fixed covariates. In addition, the concordance was utilized the Cox proportional hazards model to describe the relationship between covariates and survival time via previous studies. Since proportional hazard assumption may fail in some cases, we may replace the Cox model by alternative model such as the accelerated failure time (AFT) model. Moreover, we further extended the procedures to data with longitudinal biomarkers. We evaluated the performance of the extended methodology via simulations and demonstrated the usefulness of our procedures through three real data.
