Estimating Multiple Levels of Attribute Mastery in Diagnostic Measurement

• In Event: 18.020 - Research on Cognitive Diagnostic Modeling

Abstract

Diagnostic measurement provides fine-grained attribute-level information about examinees by classifying them with respect to each attribute as masters or non-masters. Beyond classifying examinees into this type of binary state on attributes, researchers and practitioners often want to provide examinees with multiple levels of attribute mastery, and therefore are in need of methods that allow for polytomous attribute classifications. We propose a new hierarchical approach to accommodate polytomous score reporting within each attribute. This hierarchical approach uses multiple binary attributes to represent multiple levels of attribute mastery and constrains the relationships among those attributes through specifying a hierarchical chain. Results show that the hierarchical approach with certain design features produced higher classification accuracy and reliability than the traditional polytomous approach.

Authors

• Ren Liu, University of Florida

• Anne Corinne Huggins-Manley, University of Florida

• Amber Elizabeth Benedict, University of Florida

• M. David Miller, University of Florida

• Mary T. Brownell, University of Florida

• Cameron Martin