• Corpus ID: 220364527

High-recall causal discovery for autocorrelated time series with latent confounders

  title={High-recall causal discovery for autocorrelated time series with latent confounders},
  author={Andreas Gerhardus and Jakob Runge},
We present a new method for linear and nonlinear, lagged and contemporaneous constraint-based causal discovery from observational time series in the presence of latent confounders. We show that existing causal discovery methods such as FCI and variants suffer from low recall in the autocorrelated time series case and identify low effect size of conditional independence tests as the main reason. Information-theoretical arguments show that effect size can often be increased if causal parents are… 

Figures from this paper

NTS-NOTEARS: Learning Nonparametric Temporal DAGs With Time-Series Data and Prior Knowledge
We propose a score-based DAG structure learning method for time-series data that captures linear, nonlinear, lagged and instantaneous relations among variables while ensuring acyclicity throughout
Survey and Evaluation of Causal Discovery Methods for Time Series
The main conclusions one can draw from this survey is that causal discovery in times series is an active research field in which new methods are regularly proposed, and that no family or method stands out in all situations.
Graph Neural Network based Root Cause Analysis Using Multivariate Time-series KPIs for Wireless Networks
A novel root cause analysis (RCA) framework that integrates graph neural networks (GNNs) with graph structure learning (GSL) to infer hidden dependencies from available data and showed that higher accuracy of identifying a root cause and victim nodes can be achieved when the number of nodes in an environment is increased.
A recent approach combines time-series filtering and Granger Causality
  • 2022
Inferring extended summary causal graphs from observational time series
This study addresses the problem of learning an extended summary causal graph on time series by adapting two well-known algorithms, namely PC and FCI and introduces generalizations of the causation entropy measure to any lagged or instantaneous relations.
Decoupling between ecosystem photosynthesis and transpiration: a last resort against overheating
Ecosystems are projected to face extreme high temperatures more frequently in the near future. Various biotic coping strategies exist to prevent heat stress. Controlled experiments have recently
Deep Recurrent Modelling of Granger Causality with Latent Confounding
Methods and tools for causal discovery and causal inference
This article aims at covering several methodologies that have been developed for causal discovery and causal inference, and provides a practical toolkit for interested researchers and practitioners.
Granger-Causal Link Discovery in Large Temporal Networks through Conditional Models
This paper proposes to learn time-lagged model parameters with the objective of improving recall of links, while learning to defer predictions when the overlap assumption is violated over observed time series to demonstrate a 25% increase in the area under the precision-recall curve for discovering Grangercausal links.
Quantifying causal pathways of teleconnections
The particular challenges and advantages these imply for climate science are discussed and it is argued that a systematic causal approach to statistical inference should become standard practice in the study of teleconnections.


Causal Inference on Time Series using Restricted Structural Equation Models
This work studies a class of restricted Structural Equation Models for time series that require independent residual time series, and shows empirically that when the data are causally insufficient or the model is misspecified, the method avoids incorrect answers.
On Causal Discovery from Time Series Data using FCI
From simulations, it is seen that while the FCI based approaches are in principle quite powerful for finding causal relationships in time series data, such methods are not very reliable for most practical sample sizes.
Search for Additive Nonlinear Time Series Causal Models
A feasible procedure for learning a class of non-linear time series structures, which is called additive non- linear time series, and it is shown that for data generated from stationary models of this type, two classes of conditional independence relations among time series variables and their lags can be tested efficiently and consistently using tests based on additive model regression.
Detecting and quantifying causal associations in large nonlinear time series datasets
A novel causal discovery method that flexibly combines linear or nonlinear conditional independence tests with a causal discovery algorithm to estimate causal networks from large-scale time series datasets is introduced.
A Linear Non-Gaussian Acyclic Model for Causal Discovery
This work shows how to discover the complete causal structure of continuous-valued data, under the assumptions that (a) the data generating process is linear, (b) there are no unobserved confounders, and (c) disturbance variables have non-Gaussian distributions of non-zero variances.
Causal discovery and inference: concepts and recent methodological advances
The constraint-based approach to causal discovery, which relies on the conditional independence relationships in the data, is presented, and the assumptions underlying its validity are discussed.
Causal Structure Learning from Multivariate Time Series in Settings with Unmeasured Confounding
We present constraint-based and (hybrid) score-based algorithms for causal structure learning that estimate dynamic graphical models from multivariate time series data. In contrast to previous work,
Causation, prediction, and search
The authors axiomatize the connection between causal structure and probabilistic independence, explore several varieties of causal indistinguishability, formulate a theory of manipulation, and develop asymptotically reliable procedures for searching over equivalence classes of causal models.
Discovery of Causal Models that Contain Latent Variables Through Bayesian Scoring of Independence Constraints
This work introduces a hybrid method that derives a Bayesian probability that the set of independence tests associated with a given causal model are jointly correct, and is able to score multiple causal models, which possibly contain latent variables, and output the most probable one.
Necessary and sufficient conditions for causal feature selection in time series with latent common causes
The results show that the method leads to essentially no false positives and relatively low false negative rates, even in confounded environments with non-unique lag effects, outperforming the widely used Granger causality and two more methods.