Partially and Fully Noncompensatory Response Models for Dichotomous and Polytomous Items

Area of Science:

• Psychometrics
• Educational Measurement
• Item Response Theory (IRT)

Background:

• Sympson's partially noncompensatory dichotomous response model is a key framework in IRT.
• Existing models may not fully capture complex response processes.

Purpose of the Study:

• To extend Sympson's model to ordered response data.
• To introduce and evaluate fully noncompensatory models for dichotomous and polytomous data.
• To compare the performance of partially and fully noncompensatory models.

Main Methods:

• Theoretical comparison of partially and fully noncompensatory response models.
• Monte Carlo simulations to assess parameter recovery.
• Evaluation of model fit and sampling variability.

Main Results:

• Both model types showed similar data fit when correctly specified.
• Fully noncompensatory models exhibited lower sampling variability and bias compared to partially noncompensatory models.
• Fully noncompensatory models performed better in parameter recovery.

Conclusions:

• Fully noncompensatory models are theoretically sound and empirically robust for IRT.
• These models should be considered for applications involving conjunctive response processes.
• The findings advance IRT methodology for analyzing complex response patterns.