I want to run an analysis using causal impact tool. I have one test group but multiple control groups. Can I use multiple control groups all together in one model? Eg: Y = test and A,B,C as control all together as multivariate in this tool? Any suggestions?


  • $\begingroup$ What is the difference among your controls? How can they all be controls? $\endgroup$ Feb 9, 2016 at 6:17
  • $\begingroup$ In causal statistics, causation is usually discussed relative to another potential situation. The effect of a drug is not well defined; one must estimate the effect of the drug relative to placebo or the effect of the drug relative to no treatment. Could you separately estimate effects relative to each of your control levels? $\endgroup$ Feb 9, 2016 at 6:19

1 Answer 1


Yes, absolutely. CausalImpact constructs a counterfactual to the observed post-intervention outcomes using a combination of all the control time series you enter. So in practice you almost always want at least a few control time series. Keep in mind that the model assumes all of these to be unaffected by the treatment.

The package documentation has more details:

To illustrate how the package works, we create a simple toy dataset. It consists of a response variable y and a predictor x1. Note that in practice, we’d strive for including many more predictor variables and let the model choose an appropriate subset. [...]

And further below:

Analyses may easily contain tens or hundreds of potential predictors (i.e., columns in the data function argument). Which of these were informative? We can plot the posterior probability of each predictor being included in the model using:

plot(impact$model$bsts.model, "coefficients")


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.