What is the difference between centered and non-centered parameterization?

Centered parameterization defines group effects directly relative to the population mean. Non-centered parameterization uses a 'raw' standard normal variable scaled by group variance, which is often more efficient for MCMC sampling when data is weak or group variance is small.

Can this skill help with MCMC divergences?

Yes, it includes diagnostic patterns to identify when divergences are caused by parameterization mismatches and provides the necessary code adjustments to stabilize the model.

How does partial pooling help my statistical model?

Partial pooling allows group-level estimates to 'shrink' toward the population mean based on sample size and variance, providing more robust estimates than treating groups as either completely identical or entirely independent.

Which modeling languages does this skill support?

This skill primarily provides optimized implementation patterns and best practices for Stan and JAGS.

Hierarchical Bayesian Modeling

Name: Hierarchical Bayesian Modeling
Author: choxos

bychoxos

データサイエンスとML

Implements and optimizes hierarchical Bayesian models with support for partial pooling and advanced parameterization techniques.

概要

This skill empowers Claude to architect complex multilevel and hierarchical Bayesian models, providing specialized guidance for handling nested data structures like students within schools or repeated measurements on individuals. It offers production-ready implementation patterns for Stan and JAGS, specifically focusing on solving common MCMC sampling issues. By providing templates for centered and non-centered parameterizations and partial pooling logic, the skill helps developers and data scientists navigate the 'Eight Schools' problem and minimize divergences in their statistical models.

主な機能

Generates Stan and JAGS code for nested and grouped data structures.
0 GitHub stars
Implements partial pooling logic for intelligent group-level shrinkage.
Offers templates for meta-analysis and repeated measurement study designs.
Provides centered vs. non-centered parameterization patterns to resolve sampling divergences.
Includes diagnostic guidance for tau posterior and MCMC performance issues.

ユースケース

Analyzing educational or medical datasets where subjects are nested within institutions.
Conducting meta-analyses that account for significant study-level variation.
Optimizing MCMC sampling for models with small group-level variances or sparse data.

概要

主な機能

Generates Stan and JAGS code for nested and grouped data structures.
0 GitHub stars
Implements partial pooling logic for intelligent group-level shrinkage.
Offers templates for meta-analysis and repeated measurement study designs.
Provides centered vs. non-centered parameterization patterns to resolve sampling divergences.
Includes diagnostic guidance for tau posterior and MCMC performance issues.

ユースケース

Analyzing educational or medical datasets where subjects are nested within institutions.
Conducting meta-analyses that account for significant study-level variation.
Optimizing MCMC sampling for models with small group-level variances or sparse data.