Shrinkage Regression Estimators and Their Feasibilities

« Back to list
Shrinkage Regression Estimators and Their Feasibilities
Author: Masayuki Jimichi
Specifications: ISBN  978-4862832283
142 pages
16.0 x 23.9 cm / 6.4 x 9.5 in (WxH)
Category: Academic & Science
Publisher: Kwansei Gakuin University Press
Nishinomiya, 2016
www.kwansei.ac.jp/press/
Buy now: amazon.co.jp

Synopsis

It has long been noted by specialists that if multicollinearity exists among explanatory variables in a regression model, the reduction in estimation accuracy when using a least-square estimator gives rise to several problems. This book considers the feasibility of improving estimation accuracy in this situation by using a shrinkage regression estimator. Specifically, it looks at the ordinary ridge regression estimator and generalized ridge regression estimator examined by Hoerl and Kennard (1970), a principal component regression estimator proposed by Kendall (1957), and the r-k class estimator first mentioned by Marquart (1970), respectively obtaining their root mean square error (RMSE) to calculate the degree of estimation improvement and subsequently assessing the feasibility of each as shrinkage regression estimators. The estimators are considered under both the standard model and the original (usual) model. The study offers the conclusion that, when used appropriately, shrinkage regression estimators may have advantages over least-square estimators under conditions of multicollinearity.

This book is an English-language publication.