NotesFAQContact Us
Search Tips
ERIC Number: ED198157
Record Type: RIE
Publication Date: 1980-Nov
Pages: 163
Abstractor: N/A
Reference Count: 0
The Collinearity Free and Bias Reduced Regression Estimation Project: The Theory of Normalization Ridge Regression. Report No. 2.
Bulcock, J. W.; And Others
Multicollinearity refers to the presence of highly intercorrelated independent variables in structural equation models, that is, models estimated by using techniques such as least squares regression and maximum likelihood. There is a problem of multicollinearity in both the natural and social sciences where theory formulation and estimation is in the mathematical model building tradition. The variety of ways in which analysts can cope with the condition of multicollinearity is presented. The measure of multicollinearity developed is a nonlinear transformation of the maximum variance inflation factor (Vmax) which, like the correlation coefficient, gives zero for no collinearity and one for perfect collinearity. Three ridge regression estimating methods are described and their advantages and disadvantages compared. The first three disadvantages of ridge regression stem from the use of the total mean square error (MSE) criterion. The last two chapters of the report address the theory and practice of normalization ridge regression analysis. It is shown that if a new criterion called the variance normalization criterion is used, several of the inherent disadvantages based on using the MSE criterion are correctable. Results indicate that normalization ridge regression analysis has advantages (in terms of the performance indices) over seven other methods. (Author/RL)
Publication Type: Reports - Research
Education Level: N/A
Audience: N/A
Language: English
Sponsor: Natural Sciences and Engineering Research Council, Ottawa (Ontario).
Authoring Institution: Alberta Univ., Edmonton. Dept. of Educational Foundations.
Identifiers: Multicollinearity; Normalization Regression Estimation; Ridge Regression Analysis