|
Abstract:
|
While easily estimated in the presence of censored data, the commonly used hazard ratio is generally coupled with an assumption of proportional hazards. We consider an approach for inference on clinically meaningful functionals of survival (e.g., restricted mean, quantiles) amenable to avoiding strong parametric or semiparametric assumptions and accommodating scenarios of time-varying hazard ratios. In this approach we use different models to estimate conditional survival distributions on which linear contrasts are based. The contrasts are evaluated and compared on mean squared error between approaches using nonparametric recursive partitioning, Cox's proportional hazards, and Buckley-James' linear regression with censored data. The nonparametric approach was superior when semiparametric model assumptions were violated, and had a slight loss of efficiency when such assumptions do hold.
|
