Further Studies in Estimation of Life Distribution Characteristics from Censored Data.
Abstract
The main objectives of this research have been the development of smooth nonparametric estimators of quantile functions from right-censored data and the further study of smooth density estimators from censored observations. In particular, kernel-type and 'generalized' quantile estimators have been obtained under censoring which give better estimates of percentiles of the lifetime distribution than the usual product-limit quantile estimator. Asymptotic properties of these quantile estimators have been developed, including asymptotic normality and consistency. Also, the mean-square convergence and mean squared error of the kernel quantile estimators have been investigated. A data-based procedure for selecting the bandwidth has been studied using the bootstrap, and approximate confidence intervals for the true quantile have been proposed using bootstrap estimates of the sampling distribution. Theoretical results on the optimal bandwidth selection for kernel density estimators under random right censorship have also been obtained.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1987
- Accession Number
- ADA184862
Entities
People
- William J. Padgett
Organizations
- University of South Carolina