Using a Rasch Model to Account for Guessing as a Source of Low Discrimination.

Area of Science:

• Educational Measurement
• Psychometrics
• Item Response Theory

Background:

• The three-parameter logistic (3PL) model is common for item discrimination and guessing in multiple-choice questions.
• Rasch model proponents often avoid the 3PL model due to loss of unique properties when modeling guessing.
• Existing Rasch-based methods for guessing, like omitting responses, lead to information loss and do not address variable item discrimination.

Purpose of the Study:

• To propose a method for incorporating guessing into Rasch models without sacrificing their core properties.
• To present an alternative to the 3PL model that aligns with Rasch modeling principles.
• To explore treating guessing as a factor influencing item discrimination within the Rasch framework.

Main Methods:

• Proposing a novel approach within the Rasch model framework.
• Parameterizing discrimination under specific constraints to maintain Rasch model properties.
• Conceptualizing guessing as a source of low item discrimination for standard multiple-choice items.

Main Results:

• Demonstrates a method to account for guessing using Rasch models.
• Maintains the distinctive properties and advantages of Rasch models while addressing guessing.
• Offers a viable alternative to the 3PL model for handling guessing in educational assessments.

Conclusions:

• The proposed Rasch model approach effectively accounts for guessing by linking it to item discrimination.
• This method preserves the integrity and benefits of Rasch models in educational measurement.
• Further technical considerations are noted but not detailed in this article.