TFP CausalImpact

This Python package implements an approach to estimating the causal effect of a designed intervention on a time series. For example, how many additional daily clicks were generated by an advertising campaign? Answering a question like this can be difficult when a randomized experiment is not available. The package aims to address this difficulty using a structural Bayesian time-series model to estimate how the response metric might have evolved after the intervention if the intervention had not occurred [1].

As with all approaches to causal inference on non-experimental data, valid conclusions require strong assumptions. The CausalImpact package, in particular, assumes that the outcome time series can be explained in terms of a set of control time series that were themselves not affected by the intervention. Furthermore, the relation between treated series and control series is assumed to be stable during the post-intervention period. Understanding and checking these assumptions for any given application is critical for obtaining valid conclusions.

TFP CausalImpact is a Python + TensorFlow Probability implementation of the CausalImpact R package developed at Google by Kay Brodersen and Alain Hauser. TFP CausalImpact is based on both the original R package and on a Python version https://github.com/dafiti/causalimpact developed at Dafiti by Willian Fuks. TFP CausalImpact was developed at Google by Colin Carroll, David Moore, Jacob Burnim, Kyle Loveless, and others.

This is not an officially supported Google product.

[1] Inferring causal impact using Bayesian structural time-series models. Kay H. Brodersen, Fabian Gallusser, Jim Koehler, Nicolas Remy, Steven L. Scott. Annals of Applied Statistics, vol. 9 (2015), pp. 247-274. https://research.google/pubs/pub41854/