data container

File: https://github.com/pymc-devs/pymc-examples/blob/main/examples/pymc3_howto/data_container.ipynb
Reviewers:

The sections below may still be pending. If so, the issue is still available, it simply doesn't
have specific guidance yet. Please refer to this overview of updates

Known changes needed

Changes listed in this section should all be done at some point in order to get this
notebook to a "Best Practices" state. However, these are probably not enough!
Make sure to thoroughly review the notebook and search for other updates.

General updates

• Use numpy random generator (see link above for more guidance on that)
• Use try...except when loading data with pm.get_data, see point 2 in https://github.com/pymc-devs/pymc3/wiki/PyMC3-Jupyter-Notebook-Style-Guide
• Add all relevant libraries to watermark, see point 4 in https://github.com/pymc-devs/pymc3/wiki/PyMC3-Jupyter-Notebook-Style-Guide

ArviZ related

• Fix warnings in plot_hdi
• Use return_inferencedata=True everywhere, as well as inference data everywhere.
  • See for example https://nbviewer.jupyter.org/github/pymc-devs/pymc-examples/blob/main/examples/case_studies/rugby_analytics.ipynb code cell 25 as the recommended way to get posterior and posterior predictive combined in a single inferencedata.
• Use plot_posterior in 2nd to last cell to get the reference value added automatically: https://arviz-devs.github.io/arviz/api/generated/arviz.plot_posterior.html. Note that the hdi interval and point estimate can be hidden to keep the look very similar. I'll look into generating a filled kde too, I think it's possible.

Notes

Exotic dependencies

None

Start with examples/pymc3_howto/data_container.ipynb and the linked notebook-updates overview and style guide. Review the general NumPy and pm.get_data updates, watermark dependencies, and ArviZ changes, comparing the recommended inference-data pattern in the rugby analytics notebook. Done means the listed updates are applied and the notebook no longer emits the noted plot_hdi warnings.