Builds and validates Bayesian statistical models using PyMC's probabilistic programming framework.
This skill provides a comprehensive framework for building, fitting, and validating Bayesian models within Claude Code using PyMC 5.x+. It guides users through the entire probabilistic programming workflow, including data preparation, defining weakly informative priors, and implementing hierarchical structures. The skill offers deep expertise in MCMC sampling (NUTS), Variational Inference (ADVI), and critical model diagnostics like R-hat and ESS. It also includes implementation patterns for common statistical models and rigorous model comparison techniques using LOO and WAIC, ensuring robust uncertainty quantification for complex data analysis.
主要功能
01Non-centered parameterization for hierarchical models
020 GitHub stars
03Model comparison via LOO and WAIC information criteria
04Comprehensive model diagnostics (R-hat, ESS, Divergences)
05Advanced MCMC sampling using the NUTS algorithm
06Prior and posterior predictive check workflows
使用场景
01Quantifying uncertainty in business forecasts and predictive analytics
02Analyzing multi-level or grouped data using hierarchical Bayesian structures
03Implementing robust regression models that are resistant to outliers
