A relative error estimation approach for single index model
Methodology
2016-12-12 v2
Abstract
A product relative error estimation method for single index regression model is proposed as an alternative to absolute error methods, such as the least square estimation and the least absolute deviation estimation. It is scale invariant for outcome and covariates in the model. Regression coefficients are estimated via a two-stage procedure and their statistical properties such as consistency and normality are studied. Numerical studies including simulation and a body fat example show that the proposed method performs well.
Cite
@article{arxiv.1609.01553,
title = {A relative error estimation approach for single index model},
author = {Zhanfeng Wang and Zimu Chen and Yaohua Wu},
journal= {arXiv preprint arXiv:1609.01553},
year = {2016}
}
Comments
17 pages