Finding statistical models using psychometric tests, matric results, and biographical data to predict academic success at a South African university

Abstract

Tertiary education is expensive for the individual, the family, the institution and the country. The aim of this study was to find models that could predict academic success, which is seen in this study as the completion of a specific degree in the minimum prescribed time. The reliability and construct validity of the subtests of several psychometric tests used were examined. The subtests that were found reliable and construct valid, as well as biographical data and academic history data, were used as independent variables in model fitting procedures to find models to predict academic success. The SAT 78, GSAT, SSHA, PHSF and 19 FII remain reliable instruments. The SAT 78, PHSF, and 19FII were not found to be construct valid on the sample examined. The GSAT and SSHA were found to be possibly construct valid. Four different models for each of BCom, BPharm, BA and BSc students were found to predict academic success. The CHAID procedure, which was initially used as an exploratory method selected matric results as the most significant predictor for the BCom, BPharm and BSc degree students. Matric results in combination with other predictors were selected by both the stepwise logistic regression and stepwise predictive discriminant procedures as the best predictors of academic success for all four models at the North-West University’s Potchefstroom Campus.