Ridge regression: Bonanza or beguilement?

    loading  Checking for direct PDF access through Ovid

Abstract

Points out that ridge regression is an intriguing new toy for statistical estimation theory, but that it is just that--a toy that may someday evolve into a useful if limited tool; is still too fragile to do real work. Specifically, ridge regression can indeed improve on the accuracy of traditional estimates of regression parameters if background circumstances are right. But if they are not right--and how to diagnose this remains obscure--ridge regression incurs a loss of estimational accuracy and cannot, in every case, improve on the classical procedure (ordinary least squares) for estimating predictor coefficients. (3 ref) (PsycINFO Database Record (c) 2006 APA, all rights reserved)

Related Topics

    loading  Loading Related Articles