MMM Multi-Dimensional Example

[juanitorduz]

Intro notebook to the new (experimental) MMM class.

---

📚 Documentation preview 📚: https://pymc-marketing--1521.org.readthedocs.build/en/1521/

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

juanitorduz commented on 2025-02-27T19:04:57Z
----------------------------------------------------------------

This plot looks very weird

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

juanitorduz commented on 2025-02-27T19:04:57Z
----------------------------------------------------------------

The posterior predictive distribution is very off the true values, it seems its totally missing the channel contributions.

wd60622 commented on 2025-02-28T11:36:20Z
----------------------------------------------------------------

The model seems to scale by the max value across all of the targets.

pymc-marketing/pymc_marketing/mmm/multidimensional.py

Line 829 in 0e7555c

self.scalers = self.xarray_dataset.max(dim=["date", *self.dims])

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

wd60622 commented on 2025-02-28T11:16:46Z
----------------------------------------------------------------

Line #5.        "likelihood": Prior("Normal", sigma=Prior("HalfNormal", sigma=6)),

The intercept is by geo. Should the sigma be by geo as well

`Prior("Normal", sigma=Prior("HalfNormal", sigma=6, dims="geo"), dims=("date", "geo"))

Also, sigma=6 is quite large in the scaled domain

[williambdean]

The model seems to scale by the max value across all of the targets.

pymc-marketing/pymc_marketing/mmm/multidimensional.py

Line 829 in 0e7555c

self.scalers = self.xarray_dataset.max(dim=["date", *self.dims])

---

View entire conversation on ReviewNB

[codecov]

Codecov Report

Attention: Patch coverage is 50.00000% with 4 lines in your changes missing coverage. Please review.

Project coverage is 92.87%. Comparing base (0a165fd) to head (d70ff3f).
Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc_marketing/mmm/multidimensional.py	60.00%	2 Missing ⚠️
pymc_marketing/mmm/plot.py	33.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1521      +/-   ##
==========================================
- Coverage   92.88%   92.87%   -0.02%     
==========================================
  Files          54       54              
  Lines        6270     6272       +2     
==========================================
+ Hits         5824     5825       +1     
- Misses        446      447       +1     

☔ View full report in Codecov by Sentry.
📢 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.

[juanitorduz]

The model seems to scale by the max value across all of the targets.

pymc-marketing/pymc_marketing/mmm/multidimensional.py

Line 829 in 0e7555c

self.scalers = self.xarray_dataset.max(dim=["date", *self.dims])

View entire conversation on ReviewNB

@carlosagostini is this a feature or a bug 😅 ?

[juanitorduz]

Waiting for some improvements from #1551 :)

[juanitorduz]

Thanks @carlosagostini ! Your fixes worked!

I am unblocked now :)

[williambdean]

Awesome!

[juanitorduz]

@carlosagostini @wd60622 I am having examples with the out-of-sample predictions (see a93dcf8) I think it's a coords dim update mismatch. I will try to debug but let me know if you have ideas of where this can happen 🙏

[juanitorduz]

I think 3f802a5 solves it (we might need to provide the coordinates for the time varying intercept)

[williambdean]

@carlosagostini @wd60622 I am having examples with the out-of-sample predictions (see a93dcf8) I think it's a coords dim update mismatch. I will try to debug but let me know if you have ideas of where this can happen 🙏

I had a similar fix here

pymc-marketing/pymc_marketing/mmm/multidimensional.py

Lines 1348 to 1350 in b2e5eef

	if "_target" in dataset_xarray:
	data["target_data"] = dataset_xarray._target.transpose("date", *self.dims)

In this commit: 9bb4df1

[juanitorduz]

@wd60622 @carlosagostini This one is ready for review! I think we can iterate on this notebook as we add more features. Let's merge this one (after your feedback) and engage with the users to get feedback! 🚀 Really AMAZING job you guys have done here!

[cetagostini]

This is great mate, amazing work! I'll take a look tonight 🔥

[cetagostini]

All good from my side!