The Effect of Multicollinearity on Prediction in Regression Models

• In Event: 32.067 - Multiple Linear Regression: The General Linear Model SIG Paper Session: Special Adaptations of the General Linear Model to Best Reflect the Research Question

Abstract

It is widely known that the presence of multicollinearity in a dataset can have detrimental effects on determining which predictors are responsible for the variation in the response (e.g. Pedhazur, 1982). There also exist some indication that the presence of multicollinearity does not impact one’s ability to accurately estimate/predict the value of the response variable for any specific values of the predictors (e.g. Weiss, 2012). This idea is not widely present in regression textbooks, nor is there much research literature that supports it. This study was conducted to examine this relationship in situations that varied in the number of predictors, strength of the association between the predictors and the response, sample size, and level of the multicollinearity among the predictors.

Authors

• Daniel J. Mundfrom, Eastern Kentucky University

• Michelle L DePoy Smith, Eastern Kentucky University

• Lisa W Kay, Eastern Kentucky University