[BUG]: Fix TransformedTargetRegressor update() transformer usage and transformer=None predict handling

[direkkakkar319-ops]

What does this implement/fix?

This PR fixes a few issues in TransformedTargetRegressor (skpro/regression/compose/_ttr.py) where valid usage could fail at runtime.

• _update now correctly calls self.transformer_.transform(...) instead of invoking the transformer directly.
• Added proper handling for transformer=None (the default), which now works consistently across:
  1._predict

2. _predict_quantiles
3. _predict_interval
4. _predict_proba

• Preserved DataFrame index and column structure when transforming y inside _update.

I also added targeted regression tests in skpro/regression/tests/test_ttr.py to cover:

• behavior when transformer=None for predict and predict_proba
• correct application of the transformer during update

---

Does this introduce a new dependency?

No, no new dependencies are introduced.

---

What should reviewers focus on?

Feedback would be especially helpful on:

1. correctness of transformer=None handling across prediction methods
2. _update behavior with sklearn-style transformers (transform, inverse_transform)
3. whether preserving index/column structure in the transformed y path is appropriate
4. whether the added tests sufficiently cover the fixes

---

Tests

Yes, tests have been added:

• test_ttr_without_transformer_predicts_and_returns_distribution
• test_ttr_update_applies_transformer_before_delegating

---

Additional notes

This is a bugfix-only PR. No API changes and no new dependencies.

---

PR checklist

For all contributions

• I've added myself to the [[list of contributors]
• The PR title starts with either [ENH], [MNT], [DOC], or [BUG].

For new estimators

• I've added the estimator to the API reference (N/A)
• I've added usage examples to the docstring (N/A)
• I've handled soft dependencies if any (N/A)

Local Test

Validation (local CI-equivalent checks)

• pre-commit run --files skpro/regression/compose/_ttr.py skpro/regression/tests/test_ttr.py passed.
• Local test run passed: 13406 passed, 33 skipped, 468 warnings

Evidence

[fkiraly]

Thanks for the fix!

I think the reason that this did not get caught by the tests is that the get_test_params did have no example where the regressor was capable of update, I think.

Could you add such an example?

[direkkakkar319-ops]

@fkiraly, youu are exactly right on the likely root cause.

[direkkakkar319-ops]

will update TransformedTrgetRegressor.gt_test_param to include at least one update-capable regressor

[direkkakkar319-ops]

so, i will be doing it by adding the 3rd params in the TransformedTargetRegressor.get_test_params

[direkkakkar319-ops]

Changes made

1. Added an update-capable example to get_test_params as it was asked for

params3 = {"regressor": OnlineRefit(DummyProbaRegressor()), "transformer": None}
return [params1, params2, params3]

2. made the transformer=None
   as transformer=None keeps the fixture stable while still fulfilling the goal that is to cover update-capable regressor behavior
   i did this change because first the tried the StandardScaler() but that introduced unrelated interval shape issues in generic tests

[direkkakkar319-ops]

ran the tests as precommits they are passing

[direkkakkar319-ops]

Hi, @fkiraly the changes you asked are done