Polytomous Item Explanatory Item Response Theory Models.

Area of Science:

• Psychometrics
• Educational Measurement
• Statistical Modeling

Background:

• Item response theory (IRT) models are crucial for educational and psychological assessments.
• Existing polytomous IRT models often lack detailed explanations for item difficulty variations.
• The multivariate generalized linear mixed modeling framework provides a flexible structure for complex data analysis.

Purpose of the Study:

• To investigate polytomous item explanatory IRT models using a multivariate generalized linear mixed modeling framework.
• To extend existing Rasch and partial credit models to incorporate item properties that explain difficulty.
• To compare the practical and methodological differences between two novel polytomous item explanatory approaches.

Main Methods:

• Utilized the linear logistic test model approach within a multivariate generalized linear mixed modeling framework.
• Developed the item location explanatory many-facet Rasch model and the step difficulty explanatory linear partial credit model.
• Applied these models to empirical data from the Carbon Cycle assessment and Verbal Aggression datasets.

Main Results:

• The two polytomous item explanatory models demonstrated significant methodological and practical differences.
• Differences were observed in the target difficulty parameters explained by item properties.
• Variations were found in the types of predictors used for item properties and the effects of these properties.

Conclusions:

• Item explanatory modeling offers significant potential and methodological advantages for understanding item difficulties.
• The developed models provide a more nuanced understanding of how item characteristics influence response patterns.
• These findings advance the field of psychometrics by offering richer explanatory power in IRT.