Noncompensatory MIRT For Passage-Based Tests.

Area of Science:

• Educational Measurement
• Psychometrics
• Cognitive Psychology

Background:

• Traditional models often treat item difficulty and student ability as intertwined.
• Passage-based tests present unique challenges due to local item dependence.
• Existing compensatory models may not fully capture the nuances of passage-based item responses.

Purpose of the Study:

• To introduce and evaluate the Passage-Based Noncompensatory Model (PB-NM) for binary items in passage-based tests.
• To compare the PB-NM with the commonly used bifactor model.
• To explore the relationship between difficulty and discrimination parameters in these models.

Main Methods:

• Development of the multidimensional noncompensatory Passage-Based Noncompensatory Model (PB-NM).
• Simulation study to compare the fit of PB-NM against the bifactor model.
• Application of the PB-NM to an actual reading comprehension test dataset.

Main Results:

• The PB-NM effectively accounts for passage-related local item dependence.
• The model parameterizes components in relation to difficulty, offering an alternative to discrimination parameters.
• Analysis revealed variations in the difficulty of passage-related and passage-independent components across item types.

Conclusions:

• The PB-NM provides a valuable framework for understanding item characteristics in passage-based assessments.
• It allows for a more nuanced assessment of passage comprehension and passage-independent proficiency.
• The model's application demonstrates its relevance in educational measurement research.