Using a statistical model to predict student success in Texas Woman's University mathematics program

Admission applications from high school students are received yearly by colleges but majority of the students that are being admitted eventually do not graduate. Colleges need to base their admission requirements on some specific factors to determine the students who will succeed. The purpose of this research was to utilize the data from first time incoming students who were enrolled full time and graduated from the mathematics department within five years to predict the success of future students seeking admission into the department. Data from Fall 2003 to Fall 2012 was used to build the predictive model. Success in this research is defined as the students that graduated from the mathematics program within five years of admission. Two models were developed, one from doing a forward stepwise logistic regression on all the datasets and the second was using cross validation to build a model with the training datasets and checking the effectiveness on the testing datasets. The findings were that the SAT Mathematics score is the best predictor of success. A student with an SAT Math score ≥ 590 has a probability of 0.5 of graduating within five years while those <590 are at risk of not graduating.