Estimating the efficiency of the Granger causality method for detecting unidirectional coupling in the presence of common low frequency interference

M. V. Kornilov ; I. V. Sysoev

Russian journal of nonlinear dynamics, Tome 13 (2017) no. 3, pp. 349-362.

Résumé

External low-frequency interference (including interference with a pronounced main frequency) is a common problem in measurements of complex signals, which can affect results of coupling estimation. Since it is impossible to completely remove the interference without affecting the signal itself, the question arises: what distorts the results of coupling estimation to a lesser extent: filtering the interference or ignoring it? The Granger causality (GC) method is one of the most popular approaches to the detection of directional coupling from observed signals. GC uses predictive empirical models, mostly, linear and nonlinear autoregressive models (recurrence maps). Since the method is highly parametric, its success depends primarily on the parameters of the models and on the properties of the signals. Therefore, the method has to be adapted to the data. In physiology and climatology, most signals have a pronounced time scale, so one of the most important problems is that of adapting the Granger causality method to signals with a selected time scale. The purpose of this paper is to formulate recommendations for using the Granger causality method for signals with a pronounced temporal scale in the presence of common low-frequency interference. In this paper, we restrict our attention to the case of testing for unilateral coupling and use the recommendations and criteria, developed earlier, for the effectiveness of the method. The sensitivity and specificity of the method are estimated based on surrogate time series. The testing is performed using reference systems of nonlinear dynamics and radiophysics. It is shown that the loss of sensitivity and specificity of the method decrease nonlinearly with increasing amplitude of the total interference. This dependence varies for different parameters of the method. If the power of interference is several per cent of the signal power, the best results can be achieved by an appropriate choice of parameters of the method rather than by filtering the interference. At a higher noise power, filtering is preferable.

Détail

Keywords: time series, coupling analysis, Granger causality, low frequency interference.

Bibliographie

[1] Granger C. W. J., “Investigating causal relations by econometric models and cross-spectral methods”, Econometrica, 37:3 (1969), 424–438 | DOI | Zbl

[2] Schreiber T., “Measuring information transfer”, Phys. Rev. Lett., 85:2 (2000), 461–464 | DOI | MR

[3] Baccalá L. A., Sameshima K., “Partial directed coherence: A new concept in neural structure determination”, Biol. Cybern., 84:6 (2001), 463–474 | DOI | Zbl

[4] Smirnov D. A., Sidak E. V., Bezruchko B. P., “A method for revealing coupling between oscillators with analytical assessment of statistical significance”, Tech. Phys. Lett., 39:7 (2013), 601–605 | DOI

[5] Theiler J., Eubank S., Longtin A., Galdrikian B., Farmer J. D., “Testing for nonlinearity in time series: The method of surrogate data”, Phys. D, 58:1–4 (1992), 77–94 | DOI | Zbl

[6] Sysoev I. V., Karavaev A. S., Nakonechny P. I., “Role of model nonlinearity for granger causality based coupling estimation for pathological tremor”, Izv. Vyssh. Uchebn. Zaved. Prikl. Nelin. Dinam., 18:4 (2010), 81–88 (Russian)

[7] Chen Y., Rangarajan G., Feng J., Ding M., “Analyzing multiple nonlinear time series with extended Granger causality”, Phys. Lett. A, 324:1 (2004), 26–35 | DOI | MR | Zbl

[8] Bezruchko B. P., Ponomarenko V. I., Prokhorov M. D., Smirnov D. A., Tass P. A., “Modeling nonlinear oscillatory systems and diagnostics of coupling between them using chaotic time series analysis: applications in neurophysiology”, Physics-Uspekhi, 51:3 (2008), 304–310 | DOI | DOI

[9] Marinazzo D., Pellicoro M., Stramaglia S., “Nonlinear parametric model for Granger causality of time series”, Phys. Rev. E, 73:6 (2006), 066216, 6 pp. | DOI

[10] Marinazzo D., Pellicoro M., Stramaglia S., “Kernel method for nonlinear Granger causality”, Phys. Rev. Lett., 100:14 (2008), 144103, 4 pp. | DOI

[11] Hesse W., Möller E., Arnold M., Schack B., “The use of time-variant EEG Granger causality for inspecting directed interdependencies of neural assemblies”, J. Neurosci. Methods, 124:1 (2003), 27–44 | DOI

[12] Sysoeva M., Sitnikova E., Sysoev I., Bezruchko B., van Luijtelaar G., “Application of adaptive nonlinear Granger causality: Disclosing network changes before and after absence seizure onset in a genetic rat model”, J. Neurosci. Methods, 226 (2014), 33–41 | DOI

[13] Chen Y., Bressler S. L., Ding M., “Frequency decomposition of conditional Granger causality and application to multivariate neural field potential data”, J. Neurosci. Methods, 150:2 (2006), 228–237 | DOI | MR

[14] Kornilov M. V., Medvedeva T. M., Bezruchko B. P., Sysoev I. V., “Choosing the optimal model parameters for Granger causality in application to time series with main timescale”, Chaos Solitons Fractals, 82 (2016), 11–21 | DOI | Zbl

[15] Kornilov M. V., Sysoev I. V., Bezruchko B. P., “Optimal selection of parameters of the forecasting models used for the nonlinear Granger causality method in application to the signals with a main time scales”, Nelin. Dinam., 10:3 (2014), 279–295 (Russian)

[16] Kornilov M. V., Sysoev I. V., “Investigating nonlinear Granger causality method efficiency at strong synchronization of systems”, Izv. Vyssh. Uchebn. Zaved. Prikl. Nelin. Dinam., 22:4 (2014), 66–75 (Russian)

[17] Kornilov M. V., Sysoev I. V., “Influence of the choice of the model structure for working capacity of nonlinear Granger causality approach”, Izv. Vyssh. Uchebn. Zaved. Prikl. Nelin. Dinam., 21:2 (2013), 74–88 (Russian)

[18] Sysoeva M. V., Dikanev T. V., Sysoev I. V., “Selecting time scales for empirical model construction”, Izv. Vyssh. Uchebn. Zaved. Prikl. Nelin. Dinam., 20:2 (2012), 54–62 (Russian) | Zbl

[19] Yakhno Yu. V., Molkov Ya. I., Mukhin D. N., Loskutov E. M., Feigin A. M., “Reconstruction of an evolution operator as a technique of analysis of epileptiform electric brain activity”, Izv. Vyssh. Uchebn. Zaved. Prikl. Nelin. Dinam., 19:6 (2011), 156–172 (Russian) | Zbl

[20] Packard N. H., Crutchfield J. P., Farmer J. D., Shaw R. S., “Geometry from a time series”, Phys. Rev. Lett., 45:9 (1980), 712–716 | DOI | MR

[21] Kougioumtzis D., “State space reconstruction parameters in the analysis of chaotic time series: The role of the time window length”, Phys. D, 95:1 (1996), 13–28 | DOI

[22] Winterhalder M., Schelter B., Hesse W., Schwab K., Leistritz L., Klan D., Bauer R., Timmer J., Witte H., “Comparison of linear signal processing techniques to infer directed interactions in multivariate neural systems”, Signal Process., 85:11 (2005), 2137–2160 | DOI | Zbl

[23] Nalatore H., Ding M., Rangarajan G., “Mitigating the effects of measurement noise on Granger causality”, Phys. Rev. E, 75:3 (2007), 031123, 10 pp. | DOI

[24] Nalatore H., Sasikumar N., Rangarajan G., “Effect of measurement noise on Granger causality”, Phys. Rev. E, 90:6 (2014), 062127, 9 pp. | DOI

[25] Medvedeva T. M., Paramonov F. B., Sysoev I. V., “The effect of increase of the Granger causality method sensitivity with rise of a measurement noise”, Nelin. Dinam., 11:4 (2015), 657–670 (Russian) | Zbl

[26] Smirnov D. A., Sysoev I. V., Seleznev E. P., Bezruchko B. P., “Reconstructing nonautonomous system models with discrete spectrum of external action”, Tech. Phys. Lett., 29:10 (2003), 824–827 | DOI