Cognitive diagnosis models for multiple strategies

Area of Science:

• Educational Measurement
• Psychometric Modeling
• Cognitive Science

Background:

• Cognitive diagnosis models (CDMs) are vital psychometric tools for assessing student proficiency profiles.
• Existing CDMs often assume uniform problem-solving strategies, which limits their real-world applicability.
• There is a need for more flexible models that capture individual differences in strategy use.

Purpose of the Study:

• To develop a generalized multiple-strategy cognitive diagnosis model (CDM) for dichotomous response data.
• To provide a unified framework accommodating diverse condensation rules and strategy selection methods.
• To enhance the accuracy of proficiency estimation in educational assessments.

Main Methods:

• Development of a generalized multiple-strategy CDM.
• Integration of various condensation rules (conjunctive, disjunctive, additive).
• Inclusion of different strategy selection approaches (probability-matching, over-matching, maximizing).
• Parameter estimation using marginal maximum likelihood via the expectation-maximization algorithm.

Main Results:

• Simulation studies demonstrated adequate recovery of model parameters.
• The proposed model showed robustness to certain model misspecifications.
• Analysis of real data validated the practical utility of the generalized CDM.

Conclusions:

• The generalized multiple-strategy CDM offers a more realistic approach to cognitive diagnosis.
• This model improves the estimation of student proficiency by acknowledging diverse strategies.
• The findings support the application of this advanced CDM in educational assessment contexts.