Graphs in State-Space Models for Granger Causality in Climate Science

TL;DR

This paper introduces a novel graphical state-space model approach for Granger causality analysis in climate science, leveraging advanced estimation algorithms to improve inference over traditional methods.

Contribution

It presents a new framework combining graph-based state-space models with Lasso regularisation and advanced optimization for better Granger causality inference in climate data.

Findings

Improved causality detection in climate datasets.

Enhanced model estimation using GraphEM and proximal algorithms.

Better performance over standard GC methods in experiments.

Abstract

Granger causality (GC) is often considered not an actual form of causality. Still, it is arguably the most widely used method to assess the predictability of a time series from another one. Granger causality has been widely used in many applied disciplines, from neuroscience and econometrics to Earth sciences. We revisit GC under a graphical perspective of state-space models. For that, we use GraphEM, a recently presented expectation-maximisation algorithm for estimating the linear matrix operator in the state equation of a linear-Gaussian state-space model. Lasso regularisation is included in the M-step, which is solved using a proximal splitting Douglas-Rachford algorithm. Experiments in toy examples and challenging climate problems illustrate the benefits of the proposed model and inference technique over standard Granger causality methods.