Heuristic cognitive diagnosis when the Q-matrix is unknown.

Area of Science:

• Educational Measurement
• Psychometrics
• Data Science

Background:

• Cognitive diagnosis models use Q-matrices to link test items to cognitive attributes.
• Parametric methods (EM, MCMC) for fitting these models face practical challenges.
• Non-parametric cluster analysis is an alternative but still requires the Q-matrix.

Purpose of the Study:

• To develop a Q-matrix-free method for cognitive diagnosis.
• To improve the accuracy of classifying examinees into proficiency classes.
• To address limitations of existing parametric and non-parametric approaches.

Main Methods:

• Proposed a novel approach using direct clustering of examinee item scores.
• Avoided the need for attribute sum-score profiles.
• Demonstrated the method's efficacy in classifying examinees.

Main Results:

• Direct clustering of item scores bypasses the requirement for a known Q-matrix.
• This Q-matrix-free approach leads to more accurate examinee classification.
• The findings offer a practical alternative to traditional cognitive diagnosis methods.

Conclusions:

• Clustering examinees based on item scores is a viable and more accurate alternative.
• This method overcomes the dependency on the Q-matrix, a common practical hurdle.
• The study advances cognitive diagnosis by simplifying classification procedures.