Year: 2008
Type: article
Abstract: The hazard ratio is commonly used for comparing survival distributions across groups.While easily estimated in the presence of censored data, by itself it does not allow physicians, patients, and regulatory agencies to easily judge the clinical relevance of any difference in survival across groups.We consider an estimation approach for clinically meaningful functionals of a survivor distribution (e.g., restricted mean, quantiles).In this approach we use different regression models to borrow information across sparse data than to form statistical contrasts of an estimated functional of interest.In the context of three data models, each analyzed by three different statistical models, we examine the ability to form accurate estimates.Specifically, we compare a nonparametric predictive model based on recursive partitioning of a multivariate predictor space to the two semi-parametric approaches of Cox's proportional hazards and Buckley-James' linear regression with censored data.First order linear contrasts across three estimated functionals of interest from each predictive model approach are compared using root mean squared error.In the examples covered we demonstrate that an adaptive nonparametric predictive model could prove markedly superior to the use of semiparametric predictive models.
Authors: Kyle Rudser, Michael LeBlanc, Scott S. Emerson
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W403004138