Inferring causal impact using Bayesian structural time-series models

   Abstract