---
name: pymc-bayesian-modeler
description: PyMC probabilistic programming skill for hierarchical Bayesian models in physics data analysis
allowed-tools:
  - Bash
  - Read
  - Write
  - Edit
  - Glob
  - Grep
metadata:
  specialization: physics
  domain: science
  category: data-analysis
  phase: 6
---

# PyMC Bayesian Modeler

## Purpose

Provides expert guidance on PyMC for Bayesian modeling in physics, including hierarchical models and advanced inference methods.

## Capabilities

- Probabilistic model construction
- NUTS/HMC sampling
- Variational inference
- Gaussian processes
- Model comparison (WAIC, LOO)
- Prior predictive checks

## Usage Guidelines

1. **Model Building**: Construct probabilistic models
2. **Priors**: Specify informative or weakly informative priors
3. **Sampling**: Use NUTS for efficient sampling
4. **Diagnostics**: Check convergence with trace plots and r-hat
5. **Comparison**: Compare models with information criteria

## Tools/Libraries

- PyMC
- arviz
- Theano/JAX