`CausalPy` Support

[s3alfisc]

See here for the repo: https://github.com/pymc-labs/CausalPy

We'd probably want to support all estimators available.

As with Bambi - we'd have to decide how to report bayesian inferential stats vs frequentist CIs, t-stats etc.

@drbenvincent fyi

[drbenvincent]

Let's do it. I've created an issue over on CausalPy to help coordinate anything we need to do on our side.

I'm new to this repo, so I'll have to ramp up on what needs to be done.

[s3alfisc]

I'm new to this repo, so I'll have to ramp up on what needs to be done.

It should be super easy. I'll try to set up a mock PR in the next days and then maybe you could guide me through all of the CausalPy intricacies (on which I am not an expert) / take over?

[dsliwka]

@drbenvincent, @s3alfisc : maketables now allows to support new packages in two ways:

• As before, add a custom extractor to maketables itself.
• A new "plug-in" extractor which requires no coupling between a package and maketables—your package never needs to import maketables. Simply add specific attributes and methods to your model result class, and maketables will automatically detect and use them. This is now documented here.

I have the impression that the latter approach could here be worthwhile to try as you can control in the specific estimator in CausalPy what the user can display in maketables. Any comments on how to make this more useful are highly welcome.

[s3alfisc]

The new plugin is a phenomenal idea @dsliwka! Very very cool.

[s3alfisc]

Draft PR for CausalPy over in the CausalPy repo:

pymc-labs/CausalPy#600