In time-series analysis, a process has a causal effect if other time-series are more precisely predicted given the causing process, a notion attributed to C.W.J. Granger. This concept was independently introduced by Tord Schweder in terms of composable processes. A composable process is a continuous time Markov chain with its transition probabilities satisfying certain conditions related to, or interpretable as, causality expressed by means of the component processes. These component processes have graphical representations of their mutual dependence structures, and are thus also known as Bayesian networks in continuous time. This talk discusses inference and detection of causality using composable processes. The tools are a representation of the intensity matrices for the process networks and a causality measure based on the Kullback distance as well as a few facts from the general theory of stochastic processes. This is joint work with Jonas Hallgren (KTH).
