Adding default priors for Binomial/Bernoulli families with logit link

[julianlheureux]

Hello! I am working on improving default priors.

Starting with some small steps, I did some examples of logistic regression models with logit link. Linear predictor for weakly informative priors should be within -5 and 5. So I tried to improve the prior scaler, with some conditionals looking for these cases.

Setting the intercept prior to N(0,1.5), and common terms to N(0, 1/sd(x_i)). When using categoric variables, priors will be set to N(0,1).

Doing these changes, we arrive to weakly informative priors (at least, better than the originals).

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 74.07407% with 7 lines in your changes missing coverage. Please review.

Project coverage is 89.82%. Comparing base (4cc3103) to head (c06ce1a).
Report is 6 commits behind head on main.

Files	Patch %	Lines
bambi/priors/scaler.py	74.07%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #830      +/-   ##
==========================================
- Coverage   90.01%   89.82%   -0.20%     
==========================================
  Files          47       47              
  Lines        3967     3991      +24     
==========================================
+ Hits         3571     3585      +14     
- Misses        396      406      +10     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[tomicapretto]

Hi @julianlheureux, thanks a lot for the contribution and sorry for the delay in the review.

Could you also re-run a few of the examples using the affected families so the docs are also updated? With that, I think we could merge this.

[julianlheureux]

Hello! I think I successfully made the changes you highlighted.
Please let me know if there's anything else in the code that needs improvement.

[tomicapretto]

Hi Julián, thanks for changes. I just suggested a few extra comments.

[julianlheureux]

I think the comments are appropriate! Thanks for helping make the code clearer and cleaner!

But there is a problem with the comments. There are some tests that fail becouse the line with the comment is too long. We have to make this shorter...

[tomicapretto]

@julianlheureux I see everything green now, so I assume you fixed the issue related to the comment length, right?

[julianlheureux]

@julianlheureux I see everything green now, so I assume you fixed the issue related to the comment length, right?

I think i managed to fix it. If you want, you can check the changes in the last commits I made, making those lines shorter.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB