A Note on Partial Least Squares Regression for Multicollinearity (A Comparative Study)
Moawad El-Fallah Abd El-Salam

Abstract
This paper presents and compares the partial least squares (PLS) regression as an alternative procedure for handling multicollinearity problem with two commonly used regression methods, which are ridge regression (RR) and principle component regression (PCR) .The performances of RR, PCR and PLS are compared to help and give future researchers a comprehensive view about the best procedure to handle multicollinearity problem. A Monte Carlo simulation study was used to evaluate the effectiveness of these three procedures. For comparison purposes, mean squared errors (MSE) were calculated the analysis including all simulations and calculations were done using statistical package S-Plus 2000 software. The results of this paper show that, the performances of RR are most efficient when the number of regressors is small, while the PLS is most efficient than others when the number of regressors is moderate and high.

Full Text: PDF