Out-of-sample predictions for BART intro notebook

[juanitorduz]

Out-of-sample predictions for BART intro notebook

Closes https://github.com/pymc-devs/pymc-bart/issues/48

• Add a section regarding out-of-sample predictions for the existing bikes example.
• Add a comment regarding extrapolating with BART models.

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[aloctavodia]

Thanks!

What about adding a first OOS example where you treat the problem as a regression and then a second one (the one you already wrote) showing that BART is not good at extrapolation?

[juanitorduz]

Thanks!

What about adding a first OOS example where you treat the problem as a regression and then a second one (the one you already wrote) showing that BART is not good at extrapolation?

Sure! I will add the OOS part to the first example as well. Should I keep the bikes example as it is?

[aloctavodia]

I would not change what you already added. What I would add is an example of out-of-sample predictions using the bike dataset, with the only difference being that we treat it as a regression problem, not a time-series problem.

We can keep the coal mining example unmodified.

[juanitorduz]

I added a regression-like example in https://github.com/pymc-devs/pymc-examples/pull/507/commits/ae24debc47d48e9b0c8de604d8eb7055a6e63b99

[OriolAbril]

@juanitorduz just for completeness, you have added significant piece of new code, so you should add yourself to both the authors section at the bottom (like you did) and at the :author: metadata at the beginning. Feel free to update whenever you work on another notebook