MULTICOLLINEARITY: AN OVERVIEW AND INTRODUCTION OF RIDGE PLS-SEM ESTIMATION

Abstract

Multicollinearity, or the existence of excessive correlations among (combinations of) predictor variables, is a commonly encountered phenomenon that affects (PLS-SEM) parameter estimates. This chapter provides an extensive overview of multicollinearity, its consequences, detection, and possible solutions. Critical to this overview is the explicit distinction among three types of multicollinearity: canonical structural multicollinearity, numerical multicollinearity, and common-factor multicollinearity. In addition, ridge PLS-SEM-an approach that combines the principles of ridge regression and PLS-SEM modelling-is introduced as an effective approach to mitigate the effects of canonical structural multicollinearity on estimation results.