Control group in ML model? I have on my Data Frame column with name "group" with values: control_group / campaing_group. And my question is should I use observations from "control" in Machine Learning model or only compaing_group observation use at this model ? Or as I assume in my case campaign_group is my train dataset and control_group my test dataset, so I do not have to also make train_tes_split ?
 A: it sounds more like control group didn't get the stimulus, thus the ads of the campaign and, the users in the campaign group were exposed to the ads.
I dont't believe this is a pre train/test split. But we can figure out together.
1.) What is the target of your ML model, you wanna measure, tell us the KPI
2.) If you know the KPI, tell us some descriptive statistics, what other variables you have and furthermore, how high is the mean in the control group for the target outcome and how high is the mean in the campaign group for the target outcome.
If mean in control group is lower than in campaign group, that might be an indicator, that we may talk about an ad stimulus. If it is nearly even (that depends on the scale of the target outcome, but i don't know yet, so we should wait for your answer), then it is possible th ads had no effect.
3.) How many rows are we talking about? How many rows in control and campaign, perhaps the whole shape with columns would be better.
4.) What is the level of aggregation? This could give is a hint, if we are dealing with time series or something that is more object based, e.g. monthly sales vs 100 different campaigns and their attributes.
Tell us more, and we might figure it out.
All the best,
Patrick
