本論文主要在探究如何在區間設限資料建構一致性指標。模擬研究 使用三種迴歸模型，PO 比例勝算模型、Cox 比例風險模型、AFT 加速 失敗模型以及四種常見的存活分布 Exponential、Weibull、Lognormal、 Loglogistic 去計算一致性指標。並且把 6 種可能的區間設限情況加總算 出整筆資料的一致性指標。比較不同樣本數的估計表現，並進一步研究 配適錯誤迴歸模型及存活分布之影響，最後本文所提出之估計方法應用 在乳癌相關的實際資料，根據一致性指標的估計值找出最適合用來分析 該資料的模型。;This thesis aims to investigate how to estimate a concordance index for interval-censored data. The simulation study employs three regression models—the Proportional Odds (PO) model, the Cox Proportional Hazards (PH) model, and the Accelerated Failure Time (AFT) model—as well as four commonly used survival distributions: Exponential, Weibull, Lognormal, and Loglogistic, to compute the concordance index. The six possible types of interval censoring are aggregated to obtain an overall concordance index for the entire dataset. The performance of the estimator is compared across different sample sizes, and further analysis is conducted on the effects of model misspecification in both the regression models and survival distributions. Finally, the proposed estimation method is applied to breast cancer data to identify the most appropriate model for analysis based on the estimated concordance index.
