Model Selection in Educational Assessment: Cognitive Diagnosis vs. Item Response Theory in a Medical Education Case Study

• In Event: AERA Poster Session Wednesday 11:45am
  In Poster Session: Innovations in Higher Education Assessment and Measurement

Abstract

Feedback motivates learning, but e↵ective feedback requires accurate information on students’ knowledge and progress. Two theoretical perspectives in educational measurement are latent trait models and cognitive diagnosis (CD) models. Latent trait models, like unidimensional item response theory (IRT) models, identify ability as a continuous construct and provide a single ability score for peer comparison, but lack
diagnostic information on specific skills. Conversely, CD models describe ability as a combination of discrete skills (attributes) and focus on immediate feedback on mastered and unmastered skills. Despite their potential, CD models are underused, with performance assessment still relying on latent trait models. This study evaluates CD and IRT frameworks using biochemistry exam data fitted with various CD and IRT models,
including HO-GDINA.

Authors

• Youn Seon Lim, University of Cincinnati

• Nicholas Burkhart, University of Cincinnati