Add pymc support to predictive explorer

[rohanbabbar04]

Description

• Add pymc support to predictive_explorer.
• Add parameter engine to predictive_explorer.
• Add test in test_predictive_explorer.py.
  • Update docs
    • [x] Code style is correct (follows pylint and black guidelines)
    • [x] Includes new or updated tests to cover the new feature
    • [x] New features are properly documented (with an example if appropriate)
    • [x] Includes a sample plot to visually illustrate the changes (only for plot-related functions)
    • [x] Widget states have been properly saved (only for notebooks with widgets) see for details.

[aloctavodia]

sample_prior_predictive emits a message. saying "Samplig", plus a list of the sampled variables. We should disable that message. It is generated with the logging module. We should suppress it locally with a context manager, or disable it before sampling and enable it again after sampling. We need to do this to avoid messing up calls to sample_prior_predictive outside of predictive_explorer

[aloctavodia]

It is not needed for this PR, but something to think about.

We may want to try to guess the value of "engine", for instance for a pymc model we could inspect the source and detect the regex pattern for "with pm.Model() as" where the particle "pm." is something that can vary. For Bambi it should also be simple to detect it.

We may want to keep the "engine" argument and default to "auto" or something similar, just in case our automatic detection fails

[rohanbabbar04]

sample_prior_predictive emits a message. saying "Samplig", plus a list of the sampled variables. We should disable that message. It is generated with the logging module. We should suppress it locally with a context manager, or disable it before sampling and enable it again after sampling. We need to do this to avoid messing up calls to sample_prior_predictive outside of predictive_explorer

Created a contextmanager to handle the logs generated during this step.

[rohanbabbar04]

It is not needed for this PR, but something to think about.

We may want to try to guess the value of "engine", for instance for a pymc model we could inspect the source and detect the regex pattern for "with pm.Model() as" where the particle "pm." is something that can vary. For Bambi it should also be simple to detect it.

We may want to keep the "engine" argument and default to "auto" or something similar, just in case our automatic detection fails

Sounds good.... Let me create a new issue for this

[rohanbabbar04]

I am currently keeping the sampling pymc code separated from the preliz code, like it was discussed in the jupyter notebook. Better to not disrupt the current implementation of the already existing preliz code. If required we can combine the codes later...

[aloctavodia]

agreed, we can refactor and combine later. Something really easy to combine is the plotting part, right?

[rohanbabbar04]

agreed, we can refactor and combine later. Something really easy to combine is the plotting part, right?

Yes...