Bayesian Cognitive Model Builder

Pooling decisions for hierarchical Bayesian cognitive models depend on sample size and parameter variance. A decision tree guides whether to use complete, no, or partial pooling, and when to switch between centered versus non-centered parameterization to avoid divergences.

MCMC diagnostics for Bayesian cognitive models require R-hat < 1.01, ESS > 400, zero divergences, and E-BFMI > 0.3. A reporting checklist enforces these thresholds alongside LOO/WAIC comparison and posterior predictive checks for manuscript-ready inference.

Prior selection for cognitive parameters uses distribution families matching constraints like RT > 0 or probabilities in [0,1]. Beta and LKJ choices are applied through a reproducible prior predictive workflow to ensure domain-informed constraints.

Non-centered parameterization is needed in hierarchical Bayesian cognitive models when limited samples cause pooling issues and divergences. It separates group-level parameters from individual deviations, improving MCMC convergence for cognitive science studies.

Posterior predictive checks for cognitive models involve comparing simulated data against observed outcomes using LOO/WAIC metrics. A reporting checklist guides sensitivity analysis by evaluating how prior choices influence posterior inference.

Hierarchical Bayesian cognitive modeling suits decision-making, learning, memory, and perception studies with limited samples. Partial pooling and informative priors stabilize parameter estimation across psychological parameters in these contexts.