Robust Tests for Survival Data Involving a Single Continuous Covariate
针对一般相对风险危险模型，标准Cox得分检验对协变量中的异常值敏感，本文提出了两类具有良好效率和稳健性的检验统计量，并推荐一个作为Cox得分检验的补充。
For a general relative risk hazard model, the standard testing procedure is the Cox score test. Unfortunately, it is very sensitive to outliers among the covariates. For the case of one covariate, two general classes of statistics are proposed that should have both good efficiency and good robustness properties. Members of these classes along with the Cox score statistic and some ad-hoc robust procedures are compared for power, with and without covariate contamination, for both large and small sample settings. On the basis of local efficiency and robustness consider- ations, one statistic is recommended as a supplement to the Cox score test.