Can I use this skill for model comparison?

Yes, the skill includes detailed patterns for comparing models using LOO (Leave-One-Out cross-validation) and WAIC, including Pareto-k diagnostic checks.

What version of PyMC does this skill support?

This skill is optimized for modern PyMC (version 5.x+) and follows current best practices including the use of InferenceData objects via ArviZ.

Is this skill helpful for choosing priors?

Absolutely. It includes a Distribution Selection Guide that recommends appropriate weakly informative priors for scale, location, and correlation parameters.

How does this skill handle NUTS divergences?

It provides specific troubleshooting steps for divergences, such as increasing the target_accept parameter or switching to non-centered parameterization for hierarchical models.

Does it support fast approximation for large datasets?

Yes, it includes patterns for Variational Inference (ADVI) which can be used for quick exploration or as a starting point for MCMC chains in large models.

PyMC Bayesian Modeling

Name: PyMC Bayesian Modeling
Author: Yezez9

byYezez9

0•

Data Science & ML

Builds, fits, and validates robust Bayesian statistical models using PyMC and ArviZ for advanced probabilistic programming.

This skill empowers Claude with specialized expertise in Bayesian statistical modeling using the PyMC library. It provides a standardized workflow for building complex probabilistic models, including hierarchical structures, time series, and robust regressions. Users can leverage this skill to implement MCMC sampling (NUTS), perform variational inference, conduct rigorous prior/posterior predictive checks, and diagnose sampling issues like divergences or low effective sample size. It is an essential companion for data scientists and researchers who require principled uncertainty quantification and model comparison using information criteria like LOO and WAIC.

Key Features

01Model comparison and selection using LOO and WAIC metrics via ArviZ

020 GitHub stars

03Hierarchical and multilevel modeling with non-centered parameterization support

04Comprehensive model diagnostics including R-hat, ESS, and divergence resolution

05Principled uncertainty quantification through prior and posterior predictive checks

06Automated MCMC sampling (NUTS) and Variational Inference (ADVI) configuration

Use Cases

01Analyzing grouped or multilevel data where information sharing across categories is required

02Comparing multiple competing statistical hypotheses using formal information criteria

03Estimating complex parameters with full uncertainty distributions in scientific research

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yezez9/research-agent pymc

For use in Claude.ai and ChatGPT