NotesFAQContact Us
Collection
Advanced
Search Tips
ERIC Number: ED178581
Record Type: RIE
Publication Date: 1979-Apr-9
Pages: 40
Abstractor: N/A
Reference Count: 0
ISBN: N/A
ISSN: N/A
The Variance Normalization Method of Ridge Regression Analysis.
Bulcock, J. W.; And Others
The testing of contemporary sociological theory often calls for the application of structural-equation models to data which are inherently collinear. It is shown that simple ridge regression, which is commonly used for controlling the instability of ordinary least squares regression estimates in ill-conditioned data sets, is not a legitimate approach. This is because the minimum mean square error criterion procedure is stochastic; fails to satisfy the boundary condition; and is inadmissible because of a cyclical relationship between the biasing parameter, the inflated residual sum of squares, and the ridge estimator. A new criterion called variance normalization criterion of ridge regression is proposed which is shown to resolve most of the dilemmas associated with the minimum mean square error approach. The several advantages of the new criterion are illustrated through application of a model of school learning in Ontario, Canada high schools. (Author/CTM)
Publication Type: Speeches/Meeting Papers; Reports - Research
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A
Identifiers: Canada; Multicollinearity; Ridge Regression
Note: Paper presented at the Annual Meeting of the American Educational Research Association (63rd, San Francisco, CA, April 8-12, 1979)