pymc-modeling

What problem does it solve?

This skill consolidates expert guidance for Bayesian modeling with PyMC v5+, enabling AI assistants to provide task-ready workflows, templates, and best practices rather than scattered references.

Core Features & Use Cases

• Comprehensive workflow guidance for model specification, priors, inference (nutpie, PyMC NUTS, NumPyro/JAX backends), diagnostics, and model comparison.
• Ready-to-use templates for core model families (Hierarchical/multilevel, GLMs, Gaussian processes, time series, BART, mixtures) and common patterns (coordinates/dims, non-centered parameterization, posterior predictive, and DO/observe causal inference).
• Guidance on ArviZ workflows and diagnostic standards (r_hat, ess, divergences, LOO/WAIC) and best practices for reproducibility.
• Real-world examples and short-form instructions that you can apply directly in your project today.

Quick Start

• Install PyMC v5+ and dependencies; clone or copy this skill into your local skills folder; load it in your assistant environment.
• Define a simple PyMC model (e.g., a linear regression with priors), run sampling, and inspect the inference data with ArviZ summaries.
• Use recommended workflows to assess convergence and compare alternative models.