Multiple Linear Regression and Multicollinearity: Nonlinear Methods for Analyzing Interaction Effects

• In Event: Linear Modeling: Methodology and Applications

Abstract

The purpose of this research is to explore the role of multicollinearity in MLR models, techniques used to reduce it, and ways in which multicollinearity reporting of interaction effects might be used for descriptive analysis and nonlinear modeling. Data was taken from a population of students studied over six years at an urban-serving university to explore persistence of at-risk students (P=35,239) based on 27 IVs. This paper explores the multicollinearity diagnostics matrix specifically, how researchers encounter multicollinearity in real world situations, and how multicollinearity can be addressed and explained, using nonlinear methods. This presentation also reveals how some of our commonly used models,such as enrollment management as a working example, contribute to existing marginalization of undergraduate students.

Author

• Donald L. Gilstrap, The University of Alabama