An instrument for predicting performance in introductory physics courses

Physics education research (PER) is an important sub-field of physics, with a focus on the teaching and learning of physics both at schools and universities. One important branch of PER has been the development of research-based assessments for measuring educational outcomes. While many instructors develop questions to assess student learning, research-based assessments undergo rigorous design, testing, and validation processes to facilitate objective comparisons between students and methods of instruction. Here we describe the development and refinement of a diagnostic physics exam designed to predict students' performance across all introductory physics courses. Using data collected at Auburn University, we calculated difficulty and discrimination coefficients for all test questions, as well as the overall correlation between diagnostic exam performance and final exam grade. By removing questions with poor predictive power of final exam performance, we were able to reduce the assessment to a more limited set of 17 questions. We found strong correlations (R-squared between 0.27 and 0.31) between scores on this assessment and final exam performance across all introductory courses, regardless of math level or physics content. The resulting instrument is currently being used to identify students likely to struggle in physics so that instructors are able to intervene early in the semester.