I'm using the CausalImpact package and noticed that the results are very sensitive to the choice of control groups being fed. My main questions are:
- Is there a recommended approach for how to choose the control groups?
- Is there risk of overfitting by providing too many control groups? (or is there some regularization that handles this under the hood)
- How does one determine an "optimal" set of controls? My results tend to be quite sensitive to the groups of (and # of) controls that I provide.