Hierarchical Bayesian estimation for cognitive models using Particle Metropolis within Gibbs (PMwG): A tutorial

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psychological Measurement

Background:

• Quantitative cognitive model estimation is crucial but often inefficient in psychological science.
• Hierarchical Bayesian frameworks are increasingly used for complex data analysis.
• Advanced sampling methods are needed to overcome computational challenges.

Purpose of the Study:

• To introduce the pmwg R package for efficient cognitive model estimation.
• To demonstrate the application of the Particle Metropolis within Gibbs (PMwG) algorithm.
• To facilitate the analysis of complex cognitive models and model selection.

Main Methods:

• Utilized the pmwg package in R for implementing cognitive models.
• Applied the Particle Metropolis within Gibbs (PMwG) sampling algorithm.
• Demonstrated with signal detection theory and jointly modeled tasks.

Main Results:

• The pmwg package enables efficient estimation of quantitative cognitive models.
• The tutorial covers simple and complex cognitive modeling scenarios.
• Model adequacy and selection are addressed within the framework.

Conclusions:

• The pmwg package and PMwG algorithm offer a robust and efficient solution for cognitive modeling.
• This approach can advance psychological science by enabling the resolution of previously intractable questions.
• The methods support the analysis of complex cognitive architectures and model comparison.