Experiments with Mixtures, Ill-Conditioning, and Ridge Regression

Article

St. John, Ralph C.   (1984, ASQC)   Bowling Green State University

Journal of Quality Technology    Vol. 16    No. 2
QICID: 5486    April 1984    pp. 81-96
List $10.00
Member $5.00

FOR A LIMITED TIME, ACCESS TO THIS CONTENT IS FREE!
You will need to be signed in.
New to ASQ? Register here.

Article Abstract

Experiments with mixtures require a special form of polynomial model called the canonical polynomial model. Moreover, many mixtures problems are also subject to additional constraints that often cause ill-conditioning, or collinearity. Using the eigenvalues and Variance Inflation Factors (VIF's) as measures of conditioning, we have looked at a variety of mixtures data sets. We have considered the effect on conditioning of such remedial measures as standardizing the variables, transforming to pseudocomponents, and ridge regression. In particular, ridge regression is used as a method for displaying the effects of collinearity on the regression coefficients. We conclude that since ill-conditioning is a problem in so many mixtures experiments, the mixtures practitioner should always use VIF's or eigenvalues to look for it.

Keywords

Statistics,Collinearity,Data analysis,Mixture experiments


Browse QIC Articles Chronologically:     Previous Article     Next Article

New Search

Featured advertisers





ASQ is a global community of people passionate about quality, who use the tools, their ideas and expertise to make our world work better. ASQ: The Global Voice of Quality.