A Higher-Order Cognitive Diagnosis Model with Ordinal Attributes for Dichotomous Response Data

Area of Science:

• Cognitive science
• Educational measurement
• Psychometrics

Background:

• Existing cognitive diagnosis models (CDMs) often oversimplify latent attributes as binary.
• This limitation may not accurately capture the nuances of skill acquisition and mastery.

Purpose of the Study:

• To introduce a novel higher-order cognitive diagnosis model (CDM) that incorporates ordinal attributes.
• To address the limitations of binary attribute assumptions in current CDMs for dichotomous response data.

Main Methods:

• Developed a higher-order CDM with ordinal attributes, integrating expert knowledge or empirical data regularization.
• Employed a sequential item response model for joint attribute distribution to capture sequential mastery.
• Utilized the expectation-maximization algorithm for model parameter estimation.

Main Results:

• A simulation study demonstrated the model's ability to recover parameters effectively.
• Analysis of real-world data confirmed the practical viability and applicability of the proposed ordinal CDM.

Conclusions:

• The proposed higher-order CDM with ordinal attributes offers a more sophisticated approach to cognitive diagnosis.
• This model enhances the accuracy of attribute profiling and assessment, particularly in educational contexts.