Comparison of Logistic Regression, Linear Probability, and Third-Degree Polynomial Models: Which Should a Researcher Use?

Abstract

The purpose of this study is to present a comparison of three types of regression models that could be used to analyze dichotomous criterion variables under three different data structures, and to discuss the implications of the results of those comparisons. The three types of models used were (a) logistic regression, (b) linear ordinary least squares, and (c) polynomial ordinary least squares models.