Rasch-CDM: A Combination of Rasch and Cognitive Diagnosis Models to Assess a Learning Progression

Learning progressions (LPs) have been highlighted in science education over the past ten years. LPs have the potential to guide curriculum development, instruction, and assessments. An LP has to be validated using empirical evidence, and Rasch model is one of the most broadly used measurement models to connect the empirical evidence with the hypothesized LP. However, Rasch model has limitations as it only provides students’ overall ability measures and item difficulties, offering limited information about the specific cognitive skills (i.e., attributes) needed to complete specific tasks. This study thus develops an approach that combines Rasch model and cognitive diagnosis models (CDMs) to assess students’ ability and fine-grained attributes. Students’ ability, the difficulty of individual attributes, as well as students’ attribute mastery patterns estimated from the Rasch-CDM approach can be used to validate LPs. Additionally, attribute mastery patterns of LP levels, ability distribution, and attribute difficulty can be visualized on a map, named as MGZA, showing how students progress towards more sophisticated thinking about a topic. We demonstrate this approach by validating the LP of buoyancy.