Distribution-Free Inference for Arbitrary Functionals of Survival

• Activity Number: 441
• Type: Invited
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Kyle D. Rudser*+ and Michael LeBlanc and Scott Emerson
• Companies: University of Minnesota and Fred Hutchinson Cancer Research Center and University of Washington
• Keywords: survival ; trees ; nonparametric ; restricted mean ; quantile ; hazard

Abstract

While easily estimated in the presence of censored data, the hazard does not allow the clinical relevance of differences in survival across groups to be easily judged. We consider an approach for inference on clinically meaningful functionals of a survivor distribution (e.g., restricted mean, quantiles) amenable to avoiding strong parametric or semi-parametric assumptions. In this approach we dissociate use of the hazard for estimation of the survivor function from the choice of summary measure used for comparisons. Linear contrasts are evaluated on root mean squared error and coverage between approaches using nonparametric recursive partitioning, Cox's proportional hazards, and Buckley-James' linear regression with censored data. The nonparametric approach was superior when semi-parametric model assumptions were violated and had a slight loss of efficiency when such assumptions do hold.