A polychotomous, multivariate regression model predicting performance level in a core mathematics course




Ingram, Paul Burton

Journal Title

Journal ISSN

Volume Title



Enrolling students in a mathematics course commensurate with their performance ability is of great importance to mathematics departments at all institutions of higher learning. This research will utilize pre-existing historical data from Texas Woman's University containing readily available and easily measured factors, which most institutions of higher learning will have available, to develop a predictive model that can be used to place a student in an appropriate mathematics course. Multiple logistic regression and ordinal logistic regression methods are used to construct dichotomous and polychotomous models predicting performance level in Elementary Statistics and Elementary Analysis. External validation will be conducted on different data sets than those used in model construction to evaluate how accurately the models predict performance. Both dichotomous and polychotomous models were found to have performed well, predicting within four percent for Elementary Statistics and within ten percent for Elementary Analysis.


Pages 58-60 are missing in the original document.


Education, Pure sciences, Statistics, Mathematics education