Bootstrapping Cox's Regression Model.
Abstract
Statistical inference in Cox's regression model is usually carried out using traditional (first order) large sample theory. In the spirit of earlier success stories one might try to bootstrap data in order to better assess the sampling variability of the Cox estimator. Such a bootstrap scheme is proposed in this paper. An asymptotic justification is given, showing that inference based on the bootstrap procedure is first order equivalent to the standard one. The problem of constructing more accurate moderate-sample confidence intervals is also addressed, employing second order fine-tuning of the bootstrap. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1985
- Accession Number
- ADA162443
Entities
People
- Nils L. Hjort
Organizations
- Stanford University