Expose raw fitted GLM model in report

[ablaom]

Addresses: #40

This PR add :raw_glm_model to the list of user-specifiable report items (controlled by the report_keys hyperparameter.

The default keys now includes the new :raw_glm_model key (and all other keys, as presently the case). This means that, by default, the report points to the training data, because raw GLM models do so. I'd suggest we tag this change as breaking for this reason, although technically, adding a keys, doesn't break the API.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (924d0e9) 96.72% compared to head (12c9ff5) 96.79%.

❗ Current head 12c9ff5 differs from pull request most recent head 6bba7aa. Consider uploading reports for the commit 6bba7aa to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #41      +/-   ##
==========================================
+ Coverage   96.72%   96.79%   +0.07%     
==========================================
  Files           1        1              
  Lines         183      187       +4     
==========================================
+ Hits          177      181       +4     
  Misses          6        6              

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

[rikhuijzer]

As a nitpick comment: I think it would also make sense to not have raw_glm_model enabled by default, but I don't have a strong opinion on this. There are good arguments for both sides.

Apart from that, everything looks fine to me. Thanks for implementing this Anthony!

[ablaom]

I think it would also make sense to not have raw_glm_model enabled by default

Done.

[OkonSamuel]

Addresses: #40

This PR add :raw_glm_model to the list of user-specifiable report items (controlled by the report_keys hyperparameter.

The default keys now includes the new :raw_glm_model key (and all other keys, as presently the case). This means that, by default, the report points to the training data, because raw GLM models do so. I'd suggest we tag this change as breaking for this reason, although technically, adding a keys, doesn't break the API.

Maybe we can change the name of :raw_glm_model to glm_object or something else. It's just a suggestion though.

[ablaom]

How about glm_model?

[rikhuijzer]

Thanks @ablaom and @OkonSamuel!