-
Detecting direct causality in multivariate time series: A comparative study
Authors:
Angeliki Papana,
Elsa Siggiridou,
Dimitris Kugiumtzis
Abstract:
The concept of Granger causality is increasingly being applied for the characterization of directional interactions in different applications. A multivariate framework for estimating Granger causality is essential in order to account for all the available information from multivariate time series. However, the inclusion of non-informative or non-significant variables creates estimation problems re…
▽ More
The concept of Granger causality is increasingly being applied for the characterization of directional interactions in different applications. A multivariate framework for estimating Granger causality is essential in order to account for all the available information from multivariate time series. However, the inclusion of non-informative or non-significant variables creates estimation problems related to the 'curse of dimensionality'. To deal with this issue, direct causality measures using variable selection and dimension reduction techniques have been introduced. In this comparative work, the performance of an ensemble of bivariate and multivariate causality measures in the time domain is assessed, focusing on dimension reduction causality measures. In particular, different types of high-dimensional coupled discrete systems are used (involving up to 100 variables) and the robustness of the causality measures to time series length and different noise types is examined. The results of the simulation study highlight the superiority of the dimension reduction measures, especially for high-dimensional systems.
△ Less
Submitted 2 November, 2020;
originally announced November 2020.
-
Reducing the Bias of Causality Measures
Authors:
A. Papana,
D. Kugiumtzis,
P. G. Larsson
Abstract:
Measures of the direction and strength of the interdependence between two time series are evaluated and modified in order to reduce the bias in the estimation of the measures, so that they give zero values when there is no causal effect. For this, point shuffling is employed as used in the frame of surrogate data. This correction is not specific to a particular measure and it is implemented here o…
▽ More
Measures of the direction and strength of the interdependence between two time series are evaluated and modified in order to reduce the bias in the estimation of the measures, so that they give zero values when there is no causal effect. For this, point shuffling is employed as used in the frame of surrogate data. This correction is not specific to a particular measure and it is implemented here on measures based on state space reconstruction and information measures. The performance of the causality measures and their modifications is evaluated on simulated uncoupled and coupled dynamical systems and for different settings of embedding dimension, time series length and noise level. The corrected measures, and particularly the suggested corrected transfer entropy, turn out to stabilize at the zero level in the absence of causal effect and detect correctly the direction of information flow when it is present. The measures are also evaluated on electroencephalograms (EEG) for the detection of the information flow in the brain of an epileptic patient. The performance of the measures on EEG is interpreted, in view of the results from the simulation study.
△ Less
Submitted 18 January, 2011;
originally announced January 2011.
-
Evaluation of mutual information estimators on nonlinear dynamic systems
Authors:
A. Papana,
D. Kugiumtzis
Abstract:
Mutual information is a nonlinear measure used in time series analysis in order to measure the linear and non-linear correlations at any lag $τ$. The aim of this study is to evaluate some of the most commonly used mutual information estimators, i.e. estimators based on histograms (with fixed or adaptive bin size), $k$-nearest neighbors and kernels. We assess the accuracy of the estimators by Mon…
▽ More
Mutual information is a nonlinear measure used in time series analysis in order to measure the linear and non-linear correlations at any lag $τ$. The aim of this study is to evaluate some of the most commonly used mutual information estimators, i.e. estimators based on histograms (with fixed or adaptive bin size), $k$-nearest neighbors and kernels. We assess the accuracy of the estimators by Monte-Carlo simulations on time series from nonlinear dynamical systems of varying complexity. As the true mutual information is generally unknown, we investigate the existence and rate of consistency of the estimators (convergence to a stable value with the increase of time series length), and the degree of deviation among the estimators. The results show that the $k$-nearest neighbor estimator is the most stable and less affected by the method-specific parameter.
△ Less
Submitted 12 September, 2008;
originally announced September 2008.
