Diagnostic Classification Models for Ordinal Item Responses.

Area of Science:

• Psychometrics
• Statistical modeling
• Educational measurement

Background:

• Current diagnostic classification models (DCMs) often do not fully utilize the ordered structure of response data.
• Existing polytomous DCMs can be computationally intensive and require large sample sizes.
• There is a need for more efficient and flexible DCMs for analyzing ordinal item responses.

Purpose of the Study:

• To develop and evaluate two novel diagnostic classification models (DCMs) specifically designed for ordinal item data.
• To compare the performance of the proposed DCMs against existing polytomous DCMs that ignore data order.
• To assess the utility of the proposed models in terms of parsimony, information provision, and parameter estimation.

Main Methods:

• Development of two new diagnostic classification models (DCMs) for ordinal data.
• Application of the proposed DCMs to an operational dataset.
• Comparison of the proposed DCMs with established polytomous DCMs.
• Conducting a simulation study to evaluate parameter estimation and classification accuracy.

Main Results:

• The proposed parsimonious DCMs demonstrated performance comparable to existing polytomous DCMs.
• The new models provided valuable item-level and option-level information.
• Simulation results confirmed unbiased parameter estimates and accurate individual classification.
• The proposed models are effective with smaller sample sizes compared to current methods.

Conclusions:

• The developed DCMs offer a more efficient and effective approach for analyzing ordinal item data.
• These models are particularly advantageous for small-scale testing scenarios and limited sample sizes.
• The findings support the use of these novel DCMs in psychometric and educational measurement applications.