What does stratified analysis in survival models do?

• A. Estimates a single baseline hazard for all strata.
• B. Removes censoring.
• C. Allows different baseline hazards in different groups.
• D. Allows different covariate effects across strata.

Answer: (C)

Stratified analysis in survival models allows the baseline risk to differ across groups while keeping the effect of predictors the same across those groups. In a stratified Cox model, the hazard for someone in a given stratum is modeled as h_s(t) = h0_s(t) exp(beta'X). This means each stratum can have its own baseline hazard function, capturing different background risks, but the covariate effects (the beta coefficients) are assumed to be constant across all strata. This approach is useful when the proportional hazards pattern may vary by group due to different underlying risk profiles, so you don’t force a single baseline hazard on everyone.

It doesn’t remove censoring, and it doesn’t estimate a single baseline hazard for all strata. It also doesn’t allow different covariate effects across strata—that would require interactions between the stratification factor and the covariates.