Counterfactual reasoning with vector autoregressive models

In this video article, accompanying the paper “On Counterfactual Interventions in Vector Autoregressive Models”, we consider the problem of counterfactual reasoning in a time series setting. Counterfactual reasoning allows us to explore hypothetical scenarios, in which different choices were made in the past, so that we can explore the effects of our actions. However, it is impossible to answer counterfactual questions without first having a causal model. Here we address the problem using vector autoregressive (VAR) processes. We frame the inference of a causal model as a joint regression task where for inference we use both data with and without interventions. After inferring the causal model, we exploit linearity of the VAR model to make exact predictions about the system under counterfactual interventions. Under this approach, we may measure the total effect of any hypothetical intervention in the past.