Disentangling Different Aspects of Change in Tests with the D-Diffusion Model

Area of Science:

• Psychometrics
• Cognitive Psychology
• Statistical Modeling

Background:

• Diffusion-based item response theory (IRT) models link diffusion process parameters to latent traits.
• Standard latent trait models assume trait invariance during testing, which contradicts evidence of trait change due to learning or effort.
• Previous research indicates that cognitive abilities can fluctuate during assessments.

Purpose of the Study:

• To integrate diffusion-based IRT models with latent growth curve modeling.
• To develop a flexible measurement model where latent traits can change over time within a test.
• To investigate different models of trait change (linear/quadratic, fixed/individual-specific rates).

Main Methods:

• Combined diffusion-based IRT with latent growth curve models.
• Proposed a Bayes estimator for model fitting.
• Conducted a simulation study to assess parameter recovery.
• Applied the model to visuo-spatial perspective-taking data.

Main Results:

• The integrated model allows latent traits to evolve until stability is achieved.
• Different change processes can be modeled for distinct traits.
• Simulation results indicate good parameter recovery under specific conditions.
• Demonstrated model applicability to real-world cognitive data.

Conclusions:

• The proposed diffusion-based latent growth curve model offers a powerful framework for understanding dynamic trait changes during assessments.
• This approach enhances the measurement of cognitive abilities by accounting for within-test trait fluctuations.
• The model provides insights into the processes underlying learning and effort modulation in test-takers.