Adjusting the Cumulative GPA Using Item Response Theory

In college admissions, the predictive validity of preadmissions measures such as standardized test scores and high school grades is of wide interest. These measures are most often validated against the criterion of the first‐year grade point average (GPA). However, neither the first‐year GPA nor the four‐year cumulative GPA is an adequate indicator of academic performance through four years of college. In this study, Item Response Theory (IRT) is used to develop a more reliable measure of performance, called an IRT‐based GPA, which is used to estimate the validity of traditional preadmissions information. The data are preadmissions information and course grades for the Class of 1986 at Stanford University (N = 1564). Principal factor analysis is used as a precursor to determine the dimensionality of the course data and to partition courses into approximately unidimensional subsets, each of which is scaled independently. Results show a substantial increase in predictability when the IRT‐based GPA is used instead of the usual GPA.