I am experimenting with the CausalImpact package https://google.github.io/CausalImpact/CausalImpact.html (Brodersen et al. 2015) which uses Bayesian structural time-series models.

All of their examples have many time points pre and post intervention.

In my data set, I only have 4 measurements before the intervention, and 20 after the intervention.

Is this a problem for the validity of the model? I could not find something about how many time points you should have, but for example, the very similar prophet package says you should have many. I can provide several covariates, perhaps that helps.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.