Predicting College Performance From a Bivariate Grid: Analysis and Discussion of the Grid's Practical Utility,Accuracy,and Multivariate Logic

Abstract

Studies of academic achievement at the College of Basic Studies (CBS) of Boston University have consistently shown high school grade point average (HSGPA) and the College Board Scholastic Aptitude Test—Verbal score (SAT-V) to be the most useful predictors among numerous predictor variables studied. This finding led to use of double-entry expectancy tables to study relations between performance at CBS and scores on SAT-V and HSGPA in a sample of 2,093 students. The efficiency of an 11 × 11 double-entry expectancy table, called the CBS Grid, was compared with that of multiple regression using each of three criteria: Semester 1 grade point average (GPA), Year 1 GPA, and a dichotomous variable based on the Semester 1 GPA. With all three criteria, the CBS Grid was very nearly as efficient as multiple regression. Structural factors in the data that might account for the near-equivalent performance of the two procedures are mentioned; the potential usefulness of the approach exemplified by the CBS Grid is discussed; and methods of extending this approach to include more than two predictor variables are described. In addition to combining the predictive power of SAT-V and HSGPA efficiently, the CBS Grid eliminates the unreliability inherent in combining those predictors by subjective judgment ; it avoids the technical complexity and potential ambiguity