How Early is Early Enough: Correlating Student Performance with Final Grades.

Abstract

Student retention is one of the greatest challenges facing computer science programs. Difficulties in an introductory programming class often snowball, resulting in poor student performance. Far too often, the challenges faced by such students enrolled in a first-year programming class result in dropping the major completely. In this paper, we present an analysis of 197 students over 6 semesters from 11 sections of an introductory freshman-level programming class at a private four-year liberal arts university in the southeastern United States. The goal of this research was to find the earliest point in the course assessment sequence it might be possible to predict final grade outcomes. If such points exist, targeted intervention may potentially lead to increased degree retention. Accordingly, we measured the degree of correlation between student performance on quizzes, labs, programs, and tests compared to final course grade. Overall, the results show a strong positive correlation for all four assessment modalities. These results hold significance for educators and researchers insofar as the body of computing education research is extended by evaluating the relative effectiveness of early semester subsets of each of the four categories of student data to model class outcomes. Further, early prediction of poor performers using these assessment modalities may serve as a case example in future research aimed at improving student retention rates.