An Asymptotic Theory for Logistic Regression When Some Predictors Are Measured with Error.
Abstract
This document considers a local measurement error theory for logistic regression which is applied to four different methods: ordinary logistic regression without accounting for measurement error, a functional maximum likelihood estimate, an estimate based on linearizing the logistic function and an estimator conditioned on certain appropriate sufficient statistics. This asymptotic theory includes a bias-variance trade off, which is used to construct new estimators with better asymptotic and small sample properties. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1983
- Accession Number
- ADA145673
Entities
People
- L. A. Stefanski
- Raymond J. Carroll
Organizations
- University of North Carolina at Chapel Hill