适者生存的 Cox 模型:面向时间-事件数据的枢轴变量选择
Survival of the fittest Cox model: Pivotal variable selection for time-to-event data
Maxime van Cutsem · Sylvain Sardy
中文摘要
重新审视 Cox 比例风险模型(proportional hazards model),以改进生存分析(survival analysis)中的变量选择。对偏似然(partial likelihood)做平方根变换后,正则化参数(regularization parameter)的选择具有渐近枢轴性(asymptotically pivotal):基线风险(baseline hazard)与删失率(censoring rate)精确抵消,删失在时间上的分布方式仅通过一个量级为 1/log n 的有界修正项进入。利用鞅残差(martingale residuals),无需重排设计矩阵(design matrix)即可通过 Monte Carlo 计算该惩罚水平。所得准则同时借鉴 BIC 等信息准则(information criteria)与 lasso 等惩罚回归方法(penalized regression),兼取两者之长。在模拟数据与真实数据上,该方法显著优于日常使用的最先进方法。
关键要点
- 01问题:Cox 模型变量选择中正则化参数的选择依赖基线风险与删失分布,缺乏稳定性。
- 02方法:对偏似然做平方根变换使惩罚水平渐近枢轴,并用鞅残差经 Monte Carlo 计算。
- 03结果:该准则融合 BIC 与 lasso 的优点,在模拟和真实数据上显著优于现有常用方法。
- 04局限:删失在时间上的分布仍带来量级为 1/log n 的有界修正项。
解读
尚无解读。
原始英文摘要
arXiv:2510.19374v3 Announce Type: replace Abstract: We revisit Cox's proportional hazards model to improve variable selection in survival analysis. A square-root transformation of the partial likelihood renders the selection of the regularization parameter asymptotically pivotal: the baseline hazard and the censoring rate cancel exactly, and the way censoring is spread over time only enters through a bounded correction of order $1/\log n$. Using martingale residuals, we compute this penalty level by Monte Carlo without reordering the design matrix. The resulting criterion borrows from information criteria such as BIC and from penalized regression methods such as the lasso, taking the best of both. On simulated and real data, our method substantially improves upon state-of-the-art approaches used daily.