The Effects of Q-Matrix Design on Classification Accuracy in the Hierarchical Diagnostic Classification Model

• In Event: New Developments and Applications in Cognitive Diagnosis Models

Abstract

Diagnostic classification models classify examinees at the skill level according to their mastery of latent binary attributes. These models provide feedback on the strengths and weaknesses of students on specific learning objectives with higher accuracy in a shorter test length when compared with traditional multidimensional models. A priori specifications of which attributes load on which items, expressed in the Q-matrix, precedes and supports any inference resulting from the application of the diagnostic classification model. This study investigates the effects of Q-matrix design on classification accuracy when attribute hierarchy presents. Results indicate that classification accuracy, reliability, and convergence rates improve when attributes are measured with reachable or adjacent attributes in the Q-matrix.

Authors

• Ren Liu, University of Florida

• Anne Corinne Huggins-Manley, University of Florida

• Yuxi Qiu, University of Florida

• Miao Gao, Nanjing Normal University