Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
415366 | Computational Statistics & Data Analysis | 2008 | 11 Pages |
Abstract
The logistic regression model is a popular tool for relating a binary outcome to a set of covariates. In many applications, the covariates of this model are measured with error. An approach to nonparametric logistic regression with covariate measurement error is presented. The estimate of the log-odds is formed using boosted regression trees. The algorithm uses gradient boosting to fit the trees, and their coefficients are determined using an estimating equation closely related to the likelihood score function. The method is examined using simulations.
Related Topics
Physical Sciences and Engineering
Computer Science
Computational Theory and Mathematics
Authors
Joseph Sexton, Petter Laake,