-
Notifications
You must be signed in to change notification settings - Fork 2
Description
I really enjoyed exploring your repository. Sadly, I was not able to reproduce the results and graphs shown in your example notebook in VS Code. I had to go through minor changes in the syntax of the pyproject.toml in order to be compatible with poetry (instead of pixi). I initiated the virtual environment with python 3.12.10.
After a thorough look, I think the issue could lie within the channels.py file. More specific, the generation of the channel patterns in the function _generate_seasonal_pattern. The cosine function generates negative values, this leads to the product being < 0 and the spend being set to 0.
Also, I couldn’t figure out where the base_spend value is accounted for, other than in calculating the noise_std in the _generate_linear_trend_pattern function.
This is the output generated when I run the example notebook, with seed set to 42 as in the example:
Furthermore, I am not sure why my values of my x-Social Contributions are nearly double of the values shown in the example notebook, but the shape of the time-series is identical.
I would really appreciate you looking into this, as this would help me a lot with my MMM project!