A Latent Space Diffusion Item Response Theory Model to Explore Conditional Dependence between Responses and Response...

Area of Science:

• Psychometrics
• Cognitive Science
• Educational Measurement

Background:

• Traditional measurement models assume conditional independence, where item responses only relate through latent variables.
• This assumption is often violated in practice, showing uncaptured interactions between respondents and items.
• Existing joint models of responses and response times (RTs) also struggle with these complex interactions.

Purpose of the Study:

• To investigate the existence and cognitive sources of conditional dependence in measurement.
• To propose a novel diffusion item response theory model to capture these dependencies.
• To extract diagnostic information for respondents and items using these interactions.

Main Methods:

• Developed a diffusion item response theory model integrating a latent space for information processing rate variations.
• Mapped respondents and items onto this latent space to represent conditional dependence.
• Utilized three empirical applications and a simulation study for validation and parameter recovery.

Main Results:

• The proposed model successfully maps respondents and items, with distances indicating conditional dependence.
• Empirical applications demonstrated the utility of the latent space for understanding person-item interactions and generating diagnostic feedback.
• The simulation study confirmed the model's accuracy in parameter recovery and detection of conditional dependence.

Conclusions:

• The diffusion item response theory model effectively addresses conditional dependence beyond traditional psychometric assumptions.
• This approach offers a powerful tool for personalized diagnostics and a deeper understanding of measurement processes.
• The model provides a framework for analyzing complex respondent-item interactions in various testing contexts.