Accession Number:

ADA045138

Title:

Is Ridge Regression a Panacea.

Descriptive Note:

Technical rept.,

Corporate Author:

DESMATICS INC STATE COLLEGE PA

Personal Author(s):

Report Date:

1977-09-01

Pagination or Media Count:

22.0

Abstract:

Consider the usual regression model y Xbeta epsilon where X is a matrix of full rank, beta is a vector of unknown parameters, and epsilon is a vector of random errors such that Eepsilon 0 and Var epsilon sigma squaredI. The procedure known as ridge regression has been offered as an alternative to ordinary least squares for estimating beta, particularly in those situations where severe multicollinearity exists in X. Ridge regression involves the use of a ridge estimator, which takes the form beta capk 1XX kI Xy where k or 0. The properties of ridge regression relative to those of ordinary least squares are discussed. Although ridge regression does appear to offer promise, its use as a routine analysis method is not without shortcomings. Therefore, the question in the title is answered in the negative. Author

Subject Categories:

  • Statistics and Probability

Distribution Statement:

APPROVED FOR PUBLIC RELEASE