Abstract
We consider adaptive ridge regression estimators in the general linear model with homogeneous spherically symmetric errors. A restriction on the parameter of regression is considered. We assume that all components are non negative (i.e. on the positive orthant). For this setting, we produce under general quadratic loss such estimators whose risk function dominates that of the least squares provided the number of regressors in the least fore.
A.M.S. 2017 subject classification:
Acknowledgment
Our warm thanks to the editor for reviewing our paper, who have generously given up his valuable time to the process of peer review, which led to an improved version of the paper.