A Sequential Higher Order Latent Structural Model for Hierarchical Attributes in Cognitive Diagnostic Assessments

Area of Science:

• Cognitive Diagnosis
• Psychometrics
• Educational Measurement

Background:

• Latent attribute space definition in cognitive diagnosis models commonly uses higher-order structure or attribute hierarchical structure.
• Integrating these two structures to simultaneously accommodate higher-order latent traits and hierarchical attributes has been a significant challenge.

Purpose of the Study:

• To propose a novel sequential higher-order latent structural model (LSM) that integrates both higher-order latent traits and hierarchical attributes.
• To address the limitations of existing cognitive diagnosis models in handling complex attribute structures.

Main Methods:

• Development of a sequential higher-order latent structural model (LSM).
• Incorporation of various hierarchical structures within a higher-order latent structure framework.
• Examination of the model's feasibility using simulated data and the deterministic-inputs, noisy "and" gate (DINO) model.

Main Results:

• The sequential higher-order LSM demonstrated considerable improvement in person classification accuracy compared to the conventional higher-order LSM.
• The proposed model's effectiveness was particularly evident when a specific attribute hierarchy was present.
• Successful application illustrated through an empirical example.

Conclusions:

• The sequential higher-order LSM offers a viable solution for integrating higher-order latent traits and hierarchical attributes in cognitive diagnosis.
• This advancement enhances the accuracy of person classification in diagnostic models with complex attribute structures.
• The proposed model provides a more sophisticated approach to understanding latent attribute relationships.