Robust Cox Regression
Robust Cox Proportional Hazards Regression · Also known as: Cox model with robust standard errors, sandwich-variance Cox regression, Lin-Wei robust Cox model, robust partial likelihood regression
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.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use robust Cox regression when survival data contain clustering (e.g., repeated events per subject, subjects nested in centres) or when you suspect mild misspecification such as slight non-proportionality or unmeasured shared frailty. It is a conservative safety net: if the standard model is correct the robust and model-based standard errors converge, so little is lost by using it. Do not rely on it as a substitute for a proper frailty or mixed-effects Cox model when clustering is strong, and do not use it to salvage analyses with severely non-proportional hazards — those require time-varying coefficients or stratified Cox models instead.
Strengths & limitations
- Produces valid standard errors and confidence intervals under mild model misspecification or clustering without changing the point estimates.
- Retains the semi-parametric efficiency and familiar hazard-ratio interpretation of the Cox model.
- Handles within-subject repeated events or multi-centre data with a simple cluster() option in most software.
- Minimal computational overhead — only the variance matrix is changed, not the optimization problem.
- Conservative: if the Cox model is correctly specified, robust and naive standard errors agree asymptotically.
- Requires a reasonably large sample — the sandwich estimator is asymptotically justified and can be unreliable with small n or few events (fewer than ~50 events).
- Does not fix a substantially violated proportional hazards assumption; a time-varying coefficient model or stratification is needed in that case.
- For strong clustering or shared frailty, a frailty (random-effects Cox) model is more efficient and provides frailty variance estimates that robust Cox cannot deliver.
Frequently asked
Does robust Cox regression change the hazard ratio estimates?
No. The partial-likelihood maximisation is identical; only the variance matrix (and therefore standard errors and confidence intervals) changes. Hazard ratios are the same as from the standard Cox model.
How is the cluster variable specified?
In R's survival package use the cluster() term inside coxph(). In Stata use vce(cluster clustvar). The score residuals are then summed within clusters before forming the sandwich matrix.
When is a frailty model better than robust Cox?
When the clustering is strong and you want to estimate the magnitude of between-cluster heterogeneity, or when the random effect is scientifically meaningful (e.g., genetic frailty). Robust Cox gives valid marginal inference but cannot estimate cluster-level variance.
How many events do I need for the robust estimator to be reliable?
A common rule of thumb is at least 50 events total, and at least 5–10 events per cluster on average. Below these thresholds the sandwich estimator can be notably biased downward, producing confidence intervals that are too narrow.
Should I still check the proportional hazards assumption?
Yes, always. Robust standard errors do not correct for a violated PH assumption; they only correct for mild misspecification of the covariance structure. Use Schoenfeld residuals and the cox.zph test regardless.
Sources
- 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. DOI: 10.1080/01621459.1989.10478874 ↗
- Therneau, T. M., & Grambsch, P. M. (2000). Modeling Survival Data: Extending the Cox Model. Springer. ISBN: 978-0387987784
How to cite this page
ScholarGate. (2026, June 3). Robust Cox Proportional Hazards Regression. ScholarGate. https://scholargate.app/en/statistics/robust-cox-regression
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Cox RegressionSurvival↔ compare
- Robust RegressionStatistics↔ compare
- Survival RegressionStatistics↔ compare