Learning cause-and-effect relationships from time series data

Abstract

In the first part of this talk we will give a brief introduction into the modern causal inference framework for reasoning about cause-and-effect relationships from observational data. The focus will be on the causal discovery problem, i.e., data-driven learning of qualitative causal relationships. In the second part we will introduce two recently developed algorithms for causal discovery in time series. The first of these, LPCMCI, extends the applicability of previous algorithms by specifically allowing for unobserved confounders. The second, Ensemble-PCMCI, performs causal discovery on ensembles of time series and thus allows to relax the usually made assumption of stationarity.