Build, fit, and diagnose sophisticated Bayesian statistical models using PyMC v5+ and ArviZ.
This skill equips Claude with specialized knowledge for the PyMC ecosystem, enabling the creation of robust probabilistic models with modern API patterns. It covers the entire Bayesian workflow—from defining priors and likelihoods using coordinates and dimensions to performing advanced inference with the high-performance, Rust-based nutpie sampler. Users can leverage this skill to implement hierarchical modeling, Gaussian processes, and generalized linear models while following a strict diagnostic protocol for convergence, posterior predictive checks, and model comparison via LOO-CV.
주요 기능
01Advanced model criticism with ArviZ posterior predictive checks
02Automated diagnostic workflows for R-hat, ESS, and divergence checks
03Support for non-centered parameterization in hierarchical models
04Efficient Gaussian Process implementation using HSGP patterns
05High-performance inference using the nutpie Rust-based NUTS sampler
0612 GitHub stars
사용 사례
01Conducting model comparison and selection using Leave-One-Out Cross-Validation
02Performing causal inference and uncertainty quantification in complex datasets
03Building hierarchical multilevel models for grouped or longitudinal data
