Adding Horseshoe Prior

[julianlheureux]

Hello! In this pull request, I added the horseshoe prior to the dictionary where Bambi stores specific prior distributions.

This is Carvalho's horseshoe prior, designed for the coefficients of a multiple regression model with a large number of explanatory variables, where only a few are statistically significant.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[tomicapretto]

@julianlheureux thanks for the contribution! This is something we wanted for a long time.

Before merging, it would be nice to add a test. I imagine the following would be OK as it covers the basic use cases:

• A test where we use the prior with the default arguments.
• A test where we use the prior with non-default arguments.

In both tests I would check that

• The underlying PyMC distributions are of the expected families with the expected arguments
• The sampler runs

I would not check for the absence of divergences, etc. because we can't guarantee they won't happen.

I know the PyMC part of the tests is not trivial as one has to interact with PyMC internals. So, if you want you can get started with the test and I can share a snippet showing how to test that.

[tomicapretto]

Thanks @julianlheureux!

There's something happening with the tests, but that's unrelated to the code contribution. It has to be with data loading.