EXAMINING STATISTICAL MODELS TO PREDICT ACADEMIC SUCCESS IN EARLY COLLEGE USING MIDDLE SCHOOL DATA

The study explored whether middle school students’ select academic (grade point average [GPA], Accuplacer mathematics, reading, and writing scores, admission interview scores) and non-academic characteristics (race, ethnicity, the middle school they attended, their gender, their parents’ educational level) have any predictive power with regard to their success at an early college high school. This study compared binary logistic regression (BLR) and predictive discriminant analysis (PDA) statistical models. First-year early college academic success can be predicted using BLR and only six input factors from middle school (Accuplacer math scores, interview scores, gender, and race, as three dummy variables) with a 91% accuracy.
However, a student academic success prediction model using middle school data for a student’s first year college success could not be built. This finding asserts Astin’s Student Development Theory (1993, 1999a, 1999b), the theoretical framework that guided this study, that students can grow and improve over time and educators need to focus on cultivating and developing students’ smartness through high level instruction and coaching instead of identifying and celebrating smartness by accepting only the most prospective students into colleges (Astin, 1977, 1993, 1999a, 2017, 2018).