Model Complexity in Diffusion Modeling: Benefits of Making the Model More Parsimonious

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychometrics

Background:

• The diffusion model is a prominent tool for analyzing binary decision tasks.
• It estimates cognitive parameters like information accumulation speed and decision bias.
• The model allows for trial-to-trial variability in parameters such as drift rate, starting point, and non-decision time.

Purpose of the Study:

• To investigate the influence of intertrial variability parameters on diffusion model estimations.
• To determine if simplifying the model by constraining variability improves parameter estimation accuracy.
• To assess the impact of different optimization criteria and trial numbers on model performance.

Main Methods:

• Utilized simulation studies to systematically vary parameters and assess model behavior.
• Employed data from an empirical test-retest study to evaluate real-world performance.
• Compared model fits using different optimization criteria and varying numbers of trials.

Main Results:

• Constraining intertrial variability of the drift rate and starting point to zero improved estimation accuracy.
• Less complex models yielded more reliable estimates of core psychological parameters.
• The number of trials and optimization criteria influenced the impact of variability.

Conclusions:

• Simplifying the diffusion model by fixing certain intertrial variabilities can lead to more robust parameter estimates.
• This parsimonious approach enhances the interpretability and reliability of cognitive process modeling.
• The findings offer practical guidance for applying diffusion models in psychological research.