Improved information pooling for hierarchical cognitive models through multiple and covaried regression.

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychometrics

Background:

• Cognitive process models infer underlying cognitive mechanisms from experimental data.
• Descriptive models lack process assumptions, potentially limiting predictive validity.
• Limited observations per condition can compromise process model accuracy.

Purpose of the Study:

• To develop a method for enhancing cognitive process model fitting with limited data.
• To improve parameter estimation and predictive validity in complex experimental designs.
• To directly model predictor and covariate effects on process parameters.

Main Methods:

• A hierarchical and covaried multiple regression approach was developed.
• Recurrences across conditions, participants, items, and traits were mapped to process model parameters.
• The framework allows joint modeling of multiple conditions and predictor effects.

Main Results:

• The approach facilitates parameter estimation by systematically pooling information.
• It improves parameter recovery at low observation numbers, comparable to standard methods.
• Predictor and covariate effects on process parameters can be directly modeled.

Conclusions:

• The proposed hierarchical framework enhances cognitive process modeling, particularly for multi-factor designs.
• It enables joint modeling of more conditions, improving accuracy with limited data.
• Direct modeling of predictor effects eliminates the need for post hoc analyses.