ArviZ: Exploratory analysis of Bayesian models#
ArviZ
Exploratory analysis of Bayesian models
ArviZ is a modular and flexible Python library implementing methods grounded in well-established statistical principles and providing robust, interpretable diagnostics and visualizations for Bayesian workflow.
Key Features#
Interoperability
Integrates with all major probabilistic programming libraries: PyMC, CmdStanPy, Pyro, NumPyro, and emcee.
Large Suite of Visualizations
Provides over 30 plotting functions for visualizing distributions, MCMC diagnostics, model checking, model comparison. See the gallery for examples.
State of the Art Diagnostics
Modern, theory-grounded diagnostics and statistical tools are implemented, tested and distributed through ArviZ.
Flexible Model Comparison
Includes functions for comparing models using fast approximate cross validation and brute force methods.
Built for Collaboration
Designed for flexible cross-language serialization using netCDF or Zarr formats. ArviZ also has a Julia version that uses the same data schema.
Labeled Data
Builds on top of xarray to work with labeled dimensions and coordinates.
Sponsors and Institutional Partners#
We thank these institutions for generously supporting the development and maintenance of ArviZ.
Support ArviZ#
Contributions
Contributions and issue reports are very welcome at the GitHub repository. We have a contributing guide to help you through the process. If you have any doubts, please do not hesitate to contact us on gitter.
Citation
If you use ArviZ, please cite it using 
.
See our support page for information on how to cite in BibTeX format.
Become a Sponsor
If your company or institution uses ArviZ, we encourage you to make a donation to ArviZ or to allow employees to dedicate some of their time to ArviZ.
Donate
ArviZ is a non-profit project under the NumFOCUS umbrella. To support ArviZ financially, consider donating through the NumFOCUS website.