Ridge Regression — 1980: Advances, Algorithms, and Applications

Abstract

Hoerl and Kennard showed in 1970 that regression coefficients estimated by their “ridge regression” methods have smaller mean squared error than those estimated by least squares. Presented here are (i) advances that have been made in extending the theory; (ii) algorithms for selecting the biasing parameter, k, that have been supported by simulation; and (iii) an annotated bibliography of the published papers, including a substantial number of applications to fields including chemical engineering, air pollution, geology, marketing, economics, and sociology among others.