pymc

Build a Bayesian hierarchical model by defining group-level parameters and priors using the PyMC Python API, then fit the model to grouped data using MCMC sampling to quantify uncertainty across different patient or category groupings.

Validate MCMC sampling diagnostics for Bayesian inference by analyzing R-hat, effective sample size (ESS), and divergences using ArviZ, ensuring model convergence and reliable posterior distributions before comparing model specifications.

Compare Bayesian models using LOO and WAIC by computing information criteria on fitted PyMC models to evaluate out-of-sample predictive accuracy and select the best performing model specification.

PyMC supports variational inference as a faster alternative to MCMC sampling for large datasets, approximating posterior distributions when full Markov chain Monte Carlo sampling becomes computationally prohibitive.

Perform prior and posterior predictive checks for probabilistic programming by simulating data from defined distributions and fitted models, validating that model assumptions and posterior predictions align with observed data.

Use numpy arrays and pandas DataFrames directly for Bayesian regression analysis in PyMC, passing structured data into probabilistic models to perform linear or logistic regression and quantify prediction uncertainty.