Comparing Multiple Measure Placement Models in Mathematics

Abstract

Accurate placement into an initial college mathematics course is a key step toward the successful completion of college mathematics and, eventually, a college degree. Conversely, misplacement in mathematics may lead to a reduced likelihood of course completion and degree attainment. This study investigated the ability of two placement models to predict course success in mathematics: a hierarchical placement model utilizing cut point values determined by CHAID decision trees and an algorithmic placement model utilizing the regression equations from binary logistic regression analysis. Both models used the placement measures of high school grade point average (HSGPA) and ACT-Mathematics (ACTM) scores to predict the placement of students from a large, Midwest urban/suburban community college. The accuracy of the two placement models was compared to determine which, if either, placement model more accurately placed students in their initial college mathematics course. The ability of each placement model to discern between a successful student and an unsuccessful student was also compared. The analysis showed that the two placement models perform equally well in predicting course success and in discerning between a successful and unsuccessful student. The algorithmic model tended to underplace students by placing substantially more students in the lowest mathematics course level than the hierarchical placement model with only minimal improvement in predicting success. These results, along with the likelihood that the placement rules of the hierarchical model would be more easily understood by students and advisors, make the hierarchical model more useful for college mathematics placement.