Reconstructing regime-dependent causal relationships from observational time series

Saggioro, E. ORCID: https://orcid.org/0000-0002-9543-6338, de Wiljes, J., Kretschmer, M. ORCID: https://orcid.org/0000-0002-2756-9526 and Runge, J. (2020) Reconstructing regime-dependent causal relationships from observational time series. Chaos: An Interdisciplinary Journal of Nonlinear Science, 30 (11). 113115. ISSN 1089-7682 doi: 10.1063/5.0020538

Abstract/Summary

Inferring causal relations from observational time series data is a key problem across science and engineering whenever experimental interventions are infeasible or unethical. Increasing data availability over the past few decades has spurred the development of a plethora of causal discovery methods, each addressing particular challenges of this difficult task. In this paper, we focus on an important challenge that is at the core of time series causal discovery: regime-dependent causal relations. Often dynamical systems feature transitions depending on some, often persistent, unobserved background regime, and different regimes may exhibit different causal relations. Here, we assume a persistent and discrete regime variable leading to a finite number of regimes within which we may assume stationary causal relations. To detect regime-dependent causal relations, we combine the conditional independence-based PCMCI method [based on a condition-selection step (PC) followed by the momentary conditional independence (MCI) test] with a regime learning optimization approach. PCMCI allows for causal discovery from high-dimensional and highly correlated time series. Our method, Regime-PCMCI, is evaluated on a number of numerical experiments demonstrating that it can distinguish regimes with different causal directions, time lags, and sign of causal links, as well as changes in the variables’ autocorrelation. Furthermore, Regime-PCMCI is employed to observations of El Niño Southern Oscillation and Indian rainfall, demonstrating skill also in real-world datasets. Regime-dependent non-stationarity is a ubiquitous feature of physical systems, especially prominent in atmospheric sciences. This dependence can be looked at as an intermittent change in relationships defining the dynamics of a multivariate system, each of which can be described as a time series causal network. In this work, we develop a novel algorithm to detect regime-dependent causal relations that combines the constrained-based causal discovery algorithm PCMCI with a regime assigning linear optimization algorithm. Our method, Regime-PCMCI, is evaluated on a number of numerical experiments and demonstrates high performance in detecting a variety of regime-dependent features. Finally, Regime-PCMCI is applied to observations of El Niño Southern Oscillation and Indian rainfall, demonstrating skill in detecting well-known seasonal regimes in a real-world dataset.

Item Type	Article
URI	https://reading-clone.eprints-hosting.org/id/eprint/93821
Identification Number/DOI	10.1063/5.0020538
Refereed	Yes
Divisions	Science > School of Mathematical, Physical and Computational Sciences > Department of Mathematics and Statistics Science > School of Mathematical, Physical and Computational Sciences > Department of Meteorology
Publisher	American Institute of Physics
Download/View statistics	View download statistics for this item