Fix funnel MuEffect footguns: namespaced data vars and channel-by-name API

[drbenvincent]

Summary

• Prefix DataVarMuEffect pm.Data registration as {prefix}_{var_name} so multiple effects can share dataset column names without build-time collisions
• Add MMM.scaled_channel(channel) to select scaled spend by channel name instead of a positional index
• Update the funnel notebook, tests, and incrementality fixtures to use the new APIs

Closes #2871

Test plan

• pytest tests/mmm/test_additive_effect.py::TestDataVarMuEffect
• pytest tests/mmm/test_mmm.py::TestMultidimMMMEdgeCases::test_scaled_channel_selects_by_name
• pytest tests/mmm/test_incrementality.py::TestPostFitMutationGuard
• pytest tests/mmm/test_incrementality.py::TestMediatedMuEffects::test_evaluator_discovers_the_effect_s_own_data
• pytest tests/mmm/test_additive_effect.py -k "MediaMu or ControlMu"
• Pre-commit on changed files

Made with Cursor

---

📚 Documentation preview 📚: https://pymc-marketing--2907.org.readthedocs.build/en/2907/

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

❌ Patch coverage is 38.46154% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.70%. Comparing base (53b2b77) to head (fd99868).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc_marketing/mmm/mmm.py	14.28%	6 Missing ⚠️
pymc_marketing/mmm/additive_effect.py	66.66%	2 Missing ⚠️

❗ There is a different number of reports uploaded between BASE (53b2b77) and HEAD (fd99868). Click for more details.

HEAD has 20 uploads less than BASE

Flag	BASE (53b2b77)	HEAD (fd99868)
	28	8

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #2907       +/-   ##
===========================================
- Coverage   94.84%   71.70%   -23.14%     
===========================================
  Files         110      110               
  Lines       17389    17398        +9     
===========================================
- Hits        16492    12476     -4016     
- Misses        897     4922     +4025     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[drbenvincent]

Review round 1 (Bugbot + manual)

Bugbot

No bugs flagged on branch diff.

Manual review notes

Breaking change (intentional, closes #2871)

• DataVarMuEffect now registers pm.Data as {prefix}_{var_name} instead of unprefixed dataset names.
• Custom subclasses must read inputs via model[self.model_data_name(var)] (built-ins MediaMuEffect / ControlMuEffect updated).
• Incrementality/counterfactual paths are compatible: graph discovery uses live pm.Data names (funnel_lf_budget in tests).

MMM.scaled_channel

• Selects by channel_columns index (same as previous isel(channel=0) pattern, but name-safe).
• Raises if model not built or channel missing.

Coverage added in follow-up commit

• test_scaled_channel_selects_by_name now also asserts error when model is unbuilt.
• DataVarMuEffect class docstring documents the namespacing contract.

Out of scope / follow-ups

• Open PRs Docs: build funnel MMMs from YAML #2885 / Rename funnel MuEffect classes to reflect subgraph topology #2904 still carry notebook-local isel(channel=0) until rebased onto this API.
• Untracked FUNNEL_EFFECT_SUBGRAPH_GUIDE.md not updated here.

Round 1 fixes pushed

• Clarified DataVarMuEffect docstring
• Added unbuilt-model guard test for scaled_channel

[juanitorduz]

thanks @drbenvincent ! This is great! Can you please take a look into the failing notebook test :) ?

[williambdean]

If it is actually the same data, dont think we want differnet nodes in the graph