Bugfix rscorer discrete outcome treatment

[maartenvanhooftds]

Issue #875

Changes:

• Fix for Causal Forest not passing discrete_outcome flag to RScorer. As a result, it gets initiated by it's default value False, which is unintended behaviour if we want discrete_outcome=True
• Added a test that previously failed, but now succeeds. Based on the minimal example provided here

First contribution here, please review critically for any mistakes or inconsistencies!

I see that the CI pipeline awaits approval. I think I did everything that is expected from me, so I'll mark it as ready for review. But perhaps the pipeline flags something.

[Copilot]

Pull Request Overview

This PR fixes a bug where the discrete_outcome flag was not forwarded to RScorer in CausalForestDML.tune and adds a minimal test to ensure tuning and fitting succeed when both outcome and treatment are discrete.

• Pass discrete_outcome into RScorer during tuning.
• Add a test that runs tune and fit with discrete_outcome=True and discrete_treatment=True.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
econml/dml/causal_forest.py	Included discrete_outcome=est.discrete_outcome in RScorer instantiation within tune.
econml/tests/test_dml.py	Added test_causal_forest_tune_with_discrete_outcome_and_treatment to verify no errors when both flags are true.

Comments suppressed due to low confidence (1)

econml/tests/test_dml.py:1304

• [nitpick] The new test does not include any assertions to verify the estimator's behavior. Consider adding checks—for example, assert that the internally inferred outcome dimension (est._d_y) matches the expected number of classes or that calling est.effect(X) returns an array of the correct shape—to ensure discrete flags are honored.

def test_causal_forest_tune_with_discrete_outcome_and_treatment(self):

[kbattocchi]

Thanks for the contribution, this looks great. Once our checks pass I'll merge it.