Introducing the General Polytomous Diagnosis Modeling Framework.

Area of Science:

• Educational Measurement
• Psychometrics
• Cognitive Psychology

Background:

• Cognitive diagnosis modeling has advanced significantly, but primarily for dichotomous response data.
• Existing models often fail to adequately capture the nuances of polytomous response formats.
• There is a need for flexible models that can accommodate various item response structures in cognitive diagnosis.

Purpose of the Study:

• To propose a general cognitive diagnosis model for polytomous responses (GPDM).
• To integrate the G-DINA modeling framework with an item-splitting technique for polytomous data.
• To provide a unified framework for cognitive diagnosis applicable to both dichotomous and polytomous responses.

Main Methods:

• The general polytomous diagnosis model (GPDM) was developed by combining G-DINA principles with item-splitting.
• Polytomous items were specified using a Q-matrix, similar to dichotomous items.
• Maximum marginal likelihood (MML) estimation was implemented via an Expectation-Maximization (EM) algorithm.

Main Results:

• The GPDM framework allows for linear transformations between saturated and reduced model forms.
• Model assessment and adjustment techniques from dichotomous cognitive diagnosis can be extended to polytomous data.
• Simulation studies confirmed the model's effectiveness in handling polytomous responses compared to dichotomous ones.

Conclusions:

• The proposed general polytomous diagnosis model (GPDM) offers a versatile approach to cognitive diagnosis for polytomous data.
• The model provides a unified framework, extending existing methods and facilitating broader application in educational and psychological assessments.
• Empirical evidence from simulations and real-data examples supports the practical utility and effectiveness of the GPDM.