Response Linked Censoring: Modeling and Estimation
Abstract
In time to response investigations, such as survival studies, observations are often censored. the censoring may or may not be independent of the response. A particular form of positive dependence between censoring time and death time is defined as response linked (RL) censoring. Many examples are given of response linked censoring. We present a general method for modeling and estimation to be used when both response linked and independent censoring occur. This method involves parametric modeling with (or without) covariates and handles the special cases of all response linked and all independent censoring. We illustrate in detail are response linked censoring and the method with the Standford heart transplant data. Survival data on radio tagged animals provides a second example. The effectiveness of our methods is tested with a lung cancer data set where real cancer deaths have been randomly RL censored. Several other application areas (e.g, study of reformed heroin addicts time to relapse) are mentioned. Keywords: Survival; Dependent censoring; Independent censoring; Response linked censoring; Covariates; Data analysis.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 14, 1989
- Accession Number
- ADA212739
Entities
People
- Donald R. Hoover
- Frank M. Guess
Organizations
- Stanford University