How to interpret Model Coefficients

[Snowcatcat]

Hi,
I'm looking at the model coefficients (demo) and finding that they are pretty large compared to what we normally would get.

Digital_I 39878.361
OLV_I 11370.256

I know the media variables have been transformed and do not have linear relationships with the response, but I'm not sure what would be an appropriate way to interpret these coefficients. Is there a way to find each channel's contribution at the weekly level?

Also, from the CSV file, what are roi_mean, roi_total, and mean_response?

  rn               coef          mean_spend      mean_response       roi_mean        total_spend    total_response     roi_total             solID

1: Digital_I 39878.361 337618.08 18866.197 0.055880293 4051417 319732.39 0.07891866 2_184_3
2: OLV_I 11370.256 87266.61 2221.274 0.025453888 1832599 70239.29 0.03832770 2_184_3

[gufengzhou]

Hi, first of all, you can find the schema of all csv here.

In the _alldecomp.csv, you can find the predicted value for each independent variable as time series, which is what you're asking for as "weekly contribution".

Finally, all independent variables are transformed before the fitting as you correctly acknowledged. Because the hill function we use for saturation transformation normalises all independent to 0-1, that's why beta coefficients gets bigger.

[Snowcatcat]

Thank you so much!