Logistic Regression and Decision Tree Predictive Modeling: A Fit Function Comparison (Poster 39)

• In Event: AERA Poster Session Sunday 9:50 a.m.
  In Poster Session: Predictive Models With Cross-Validation

Abstract

Logistic Regression and Decision Tree predictive modeling are compared using the same set of independent variables with a dependent variable binary outcome. The fit function used was the classification accuracy in the confusion matrix. If an effective set of independent variables are not chosen for logistic regression or Decision Tree analyses, then the fit function (classification accuracy) would not yield the best accurate results. The independent variables for logistic regression in a sample data set, training data set, and test data set are compared to the variable information gain values in the corresponding data sets for Decision Tree. A Principal Components factor analysis provided important preliminary information about which independent variables would be the most important in the analyses.

Authors

• Randall E. Schumacker, University of Alabama

• Todd Sherron, Texas State University