Causal network reconstruction from time series: From theoretical assumptions to practical estimation.

By Jakob Runge. Chaos, 2018

439 citations recorded by the source corpus.

Causal network reconstruction from time series is an emerging topic in many fields of science. Beyond inferring directionality between two time series, the goal of causal network reconstruction or causal discovery is to distinguish direct from indirect dependencies and common drivers among multiple time series. Here, the problem of inferring causal networks including time lags from multivariate time series is recapitulated from the underlying causal assumptions to practical estimation problems. Each aspect is illustrated with simple examples including unobserved variables, sampling issues, determinism, stationarity, nonlinearity, measurement error, and significance testing. The effects of dynamical noise, autocorrelation, and high dimensionality are highlighted in comparison studies of common causal reconstruction methods. Finally, method performance evaluation approaches and criteria are suggested. The article is intended to briefly review and accessibly illustrate the foundations and practical problems of time series-based causal discovery and stimulate further methodological developments.

Subfields: Artificial Intelligence, Cognitive Neuroscience, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/0e6d57e635ebf616ae8f039aa446dfe297f6f1a5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Dino Sejdinovic
2. Marlene Kretschmer
3. Seth R. Flaxman
4. Peer Nowack
5. Jakob Runge
6. Bharath K. Sriperumbudur
7. Bernhard Schölkopf
8. Arthur Gretton

In Cognitive Neuroscience

1. Marlene Kretschmer
2. Dino Sejdinovic
3. Seth R. Flaxman
4. Peer Nowack
5. Jakob Runge

In Molecular Biology

1. Marlene Kretschmer
2. Dino Sejdinovic
3. Peer Nowack
4. Seth R. Flaxman
5. Jakob Runge

Citations in the corpus

• Detecting and quantifying causal associations in large nonlinear time series datasets (2017, positive): “If, however, the causal dependency on a common nonstationarity is unknown, the resulting causal networks can contain spurious links (34).”
• Discovering contemporaneous and lagged causal relations in autocorrelated nonlinear time series datasets (2020, neutral): “…of utilizing a wide and growing class of CI tests, from linear partial correlation (ParCorr) and non-parametric residual-based approaches [Ramsey, 2014, Runge et al., 2019b] to Kernel measures [Zhang et al., 2011], tests based on conditional mutual information [Runge, 2018b], and neural…”