In this talk, we will study Granger causality in the context of wide-sense stationary time series, where our focus is on the topological aspects of the underlying causality graph. We examine sufficient conditions (in particular, the notion of a “strongly causal" graph topology) under which Granger causality satisfies certain intuitions (particularly transitivity) about how causation should "flow" through the graph and moreover we show that in this case the true causality graph can be recovered via pairwise causality testing alone. Examples are provided from the gene regulatory network literature suggesting that our concept of a strongly causal graph may in fact arise naturally in some application areas. Finally, we study simulation evidence that efficiency gains (both statistical and computational) can be obtained (in comparison to popular LASSO-type algorithms) when these structural assumptions are met.
Bio: Ryan J. Kinnear is a second year PhD student at the University of Waterloo studying under the supervision of Ravi R. Mazumdar. Obtaining an MASc under the same supervisor in 2017, his work has focused on time series analysis and Granger causality, as well as connections to sparse machine learning methods. After a brief stint working in the financial industry he is now undertaking research in dynamic real time bidding auctions.
Welcome to everyone!