Maximizing Substantive Input From Content Experts With the pG-DINA Model

• In Event: Understanding the Cognitive Processes Tapped by Complex Assessments

Abstract

Even though some cognitive diagnosis models (CDMs) that can accommodate polytomous attributes have been developed (e.g., General Diagnostic Model), the application of such models is still uncommon in practice. To expand such application, this research proposes the pG-DINA model, a CDM for polytomous attributes that can incorporate substantive input from content experts. This new model, which is based the G-DINA framework, involves a modified Q-matrix to allow more input from experts. We evaluated the viability of the pG-DINA model by examining how well its parameters can be estimated under various simulated conditions. We also compared the person classification accuracy of the pG-DINA model and a dichotomous model modified for polytomous attributes.

Authors

• Jinsong Chen, Rutgers University

• Jimmy de la Torre, Rutgers University