Mining Student Misconceptions from Pre- and Post-Testing Data

We analyze results from paired preand post-instruction administration of the Mechanics Baseline Test to 2238 students in introductory mechanics classes. We investigate pairs of specific wrong answers given with unusual frequency by students on the pretest. We also identify transitions between preand post-test answers on the same question which elucidate student learning due to instruction. We define criteria for excess transitions above a random response model. Some common transitions are found to be associated specifically with students within a particular range of skills. Further, transitions from preto post-test revealed that incorrect pretest answers that were frequently repeated on the posttest often correspond to known misconceptions from physics or math. Thus, our data mining techniques can elucidate common student misunderstandings of mechanics concepts and how instruction affects these misunderstandings. This opens the way for finding improved interventions for specific misunderstandings revealed by analyzing results from preand post conceptual tests.