Department of Biostatistics, Epidemiology and Environmental Health Sciences, Jiann-Ping Hsu College of Public Health, Georgia Southern University, Statesboro, Georgia, United States of America.
PLoS One. 2023 Apr 26;18(4):e0278700. doi: 10.1371/journal.pone.0278700. eCollection 2023.
Cox's proportional hazards model (PH) is an acceptable model for survival data analysis. This work investigates PH models' performance under different efficient sampling schemes for analyzing time to event data (survival data). We will compare a modified Extreme, and Double Extreme Ranked Set Sampling (ERSS, and DERSS) schemes with a simple random sampling scheme. Observations are assumed to be selected based on an easy-to-evaluate baseline available variable associated with the survival time. Through intensive simulations, we show that these modified approaches (ERSS and DERSS) provide more powerful testing procedures and more efficient estimates of hazard ratio than those based on simple random sampling (SRS). We also showed theoretically that Fisher's information for DERSS is higher than that of ERSS, and ERSS is higher than SRS. We used the SEER Incidence Data for illustration. Our proposed methods are cost saving sampling schemes.
考克斯比例风险模型（PH）是一种可接受的生存数据分析模型。本研究探讨了 PH 模型在分析事件时间数据（生存数据）的不同有效抽样方案下的性能。我们将比较一种改进的极端值和双重极端值排序集抽样（ERSS 和 DERSS）方案与简单随机抽样方案。假设观测值是根据与生存时间相关的易于评估的基线可用变量选择的。通过深入的模拟，我们表明这些改进的方法（ERSS 和 DERSS）提供了更强大的检验程序和更有效的风险比估计，而不是基于简单随机抽样（SRS）的方法。我们还从理论上证明了 DERSS 的费舍尔信息量高于 ERSS，而 ERSS 高于 SRS。我们使用 SEER 发病率数据进行说明。我们提出的方法是节省成本的抽样方案。
