Ridge trace as a boost to ridge regression estimate in the presence of multicollinearity

Abstract

Multicollinearity often causes a huge interpretative problem in linear regression  analysis. The ridge estimator is not generally accepted as a vital alternative to the  ordinary least squares (OLS) estimator because it depends on unknown parameters. In any specific application of ridges regression, there is no guarantee that the sample  estimate is a member of the class of more accurate estimates. This paper therefore  reveals the importance of ridge trace to boost ridge regression estimate in the presence of multicollinearity. We observed from our sample analysis that the use of ridge trace produced a model close to the principal component regression which was used as a check model in solution to an ill-conditioned regression.

Key words: Multicollinearity, Ridge Regression, Shrinkage Parameter, Ridge Trace