方法对比

	稳健 Cox 回归	稳健回归	生存回归
领域	统计学	统计学	统计学
方法族	Regression model	Regression model	Regression model
起源年份	1989	1964	1980s
提出者	Lin & Wei	Peter J. Huber (M-estimation, 1964); Frank Hampel (influence function, 1974)	Kalbfleisch & Prentice; Cox & Oakes
类型	Semi-parametric survival regression with robust variance	Regression with outlier resistance	Parametric survival model
开创性文献	Lin, D. Y., & Wei, L. J. (1989). The robust inference for the Cox proportional hazards model. Journal of the American Statistical Association, 84(408), 1074–1078.	Huber, P. J. (1964). Robust estimation of a location parameter. The Annals of Mathematical Statistics, 35(1), 73–101.	Kalbfleisch, J. D., & Prentice, R. L. (2002). The Statistical Analysis of Failure Time Data (2nd ed.). Wiley. ISBN: 978-0471363576
别名	Cox model with robust standard errors, sandwich-variance Cox regression, Lin-Wei robust Cox model, robust partial likelihood regression	M-estimation regression, robust linear regression, outlier-resistant regression, MM-estimation	accelerated failure time model, AFT model, parametric survival model, time-to-event regression
相关	3	6	3
摘要	Robust Cox regression fits the standard Cox proportional hazards model but replaces the model-based variance estimate with a sandwich (Huber-White) estimator. This yields valid standard errors and confidence intervals even when observations are clustered, the independence assumption is mildly violated, or the working model is slightly misspecified, without discarding the familiar hazard-ratio interpretation.	Robust regression estimates the linear relationship between a continuous outcome and predictors while sharply reducing the influence of outliers and leverage points. Unlike OLS, which is highly sensitive to extreme observations, robust methods assign down-weighted influence to atypical data points, producing coefficient estimates that remain stable even when a fraction of the data is contaminated or non-normally distributed.	Survival regression models the time until an event occurs — such as death, failure, or relapse — as a function of covariates. Unlike ordinary regression, it properly accounts for censored observations (cases where the event had not yet occurred at the end of follow-up) by specifying a parametric distribution for the survival time and estimating covariate effects via maximum likelihood.