A review of standardization and centering techniques in a multicollinear regression model

Abstract

The issue of multicollinearity is well published but the available methods for multicollinearity correction is still debated. In this paper, we review the strength or performance of standardization and centering techniques in reducing multicollinearity problem in both linear and non-linear regression model using the variance inflation factor (VIF). A Monte Carlo simulation was carried out to show the precise effects of mean centering and standardization on both individual correlation coefficients as well as overall linear and non-linear model indices. Our findings reveal that use of centering and standardization are not very effective under severe collinearity. It is therefore hoped that practicing researchers will cautiously incorporate these diagnostics into their analyses.

Keywords: multicollinearity; variance inflation factor; standardization; centering, uncentered