Abstract
The purpose of this study was to determine the extent to which performance (outcomes) on the mathematics subtest of the College-Level Academic Skills Test (CLAST) could be predicted by a University of Miami diagnostic test (D-CLAST) and other variables, including high school percentile rank, ACT or SAT scores, freshman-level mathematics courses and grades, cumulative college grade point average (CGPA), and total college credits. The sample for this study consisted of 1,363 degree-seeking undergraduates at the University of Miami who took the CLAST for the first time between October 1992 and June 1994. Although passing rates on the CLAST at the University of Miami have ranked among the highest in the state, the consequences for students who fail have raised important concerns.Criterion-related validation of the independent variables was investigated using the Pearson product-moment correlation coefficient and multiple regression techniques. These were used to predict the CLAST mathematics subtest scaled score. Two-group discriminant function analysis was applied to assess whether a combination of the independent variables could predict a pass or fail outcome on the CLAST. The chi-square test of independence was used to determine whether a relationship existed between selected nominal variables and a pass/fail CLAST result.Of all independent variables in this study, the D-CLAST total score exhibited the strongest relationship with the CLAST mathematics subtest. Among the broad skills tested on the D-CLAST, algebra was the best predictor. SAT and ACT mathematics scores also were found to be good predictors. Students who had completed at least one freshman-level mathematics course before taking the CLAST (particularly a calculus course) were more likely to pass the CLAST than those who had not taken any math courses.The regression equation containing D-CLAST scores, SAT math scores, and CGPA contributed the most variance to the prediction of the CLAST mathematics score, The combination of D-CLAST scores, SAT math scores, and number of mathematics courses taken was the most accurate discriminant model found.