ABSTRACT
We propose a method that detects the true direction of time series, by fitting an autoregressive moving average model to the data. Whenever the noise is independent of the previous samples for one ordering of the observations, but dependent for the opposite ordering, we infer the former direction to be the true one. We prove that our method works in the population case as long as the noise of the process is not normally distributed (for the latter case, the direction is not identifiable). A new and important implication of our result is that it confirms a fundamental conjecture in causal reasoning --- if after regression the noise is independent of signal for one direction and dependent for the other, then the former represents the true causal direction --- in the case of time series. We test our approach on two types of data: simulated data sets conforming to our modeling assumptions, and real world EEG time series. Our method makes a decision for a significant fraction of both data sets, and these decisions are mostly correct. For real world data, our approach outperforms alternative solutions to the problem of time direction recovery.
- Balian, R. (1992). From microphysics to macrophysics. Springer.Google Scholar
- Breidt, F. J., & Davis, R. A. (1991). Time-reversibility, identifiability and independence of innovations for stationary time series. Journal of Time Series Analysis, 13(5), 379--390.Google Scholar
- Brockwell, P. J., & Davis, R. A. (1991). Time Series: Theory and Methods. Springer, 2 edn. Google ScholarDigital Library
- Darmois, G. (1953). Analyse générale des liaisons stochastiques. Rev. Inst. Internationale Statist., 21, 2--8.Google ScholarCross Ref
- Diks, C., van Houwelingen, J., Takens, F., & DeGoede, J. (1995). Reversibility as a criterion for discriminating time series. Physics Letters A, 15, 221--228.Google ScholarCross Ref
- EEGdata (2008). This data set (experiment 4a, subject 3) can be downloaded after registration. Website, 3.6.2008, 3:51pm. http://ida.first.fraunhofer.de/projects/bci/competition_iii/#datasets.Google Scholar
- Gretton, A., Bousquet, O., Smola, A., & Schöölkopf, B. (2005). Measuring statistical dependence with Hilbert-Schmidt norms. In ALT, 63--78. Springer-Verlag. Google ScholarDigital Library
- Gretton, A., Fukumizu, K., Teo, C.-H., Song, L., Schölkopf, B., & Smola, A. (2008). A kernel statistical test of independence. In Advances in Neural Information Processing Systems 20, 585--592. Cambridge, MA: MIT Press.Google Scholar
- Hawking, S. (1995). A brief history of time. Bantam.Google Scholar
- Hoyer, P. O., Janzing, D., Mooij, J., Peters, J., & Schöölkopf, B. (2009). Nonlinear causal discovery with additive noise models. Proceedings of the Twenty-Second Annual Conference on Neural Information Processing Systems (NIPS).Google Scholar
- Janzing, D. (2007). On causally asymmetric versions of Occam's Razor and their relation to thermodynamics. http://arxiv.org/abs/0708.3411.Google Scholar
- Jarque, C. M., & Bera, A. K. (1987). A test for normality of observations and regression residuals. International Statistical Review, 55 (2), 163--172.Google ScholarCross Ref
- Kankainen, A. (1995). Consistent testing of total independence based on the empirical characteristic function. PhD Thesis, University of Jyväskylä.Google Scholar
- Mamai, L. V. (1963). On the theory of characteristic functions. Select. Transl. Math. Stat. Probab., 4, 153--170.Google Scholar
- Mandelbrot, B. (1967). On the distribution of stock price differences. Operations Research, 15 (6), 1057--1062.Google ScholarDigital Library
- Pearlmutter, B. A., & Parra, L. C. (1997). Maximum likelihood blind source separation: A context-sensitive generalization of ica. In Advances in Neural Information Processing Systems 9, 613--619. MIT Press.Google Scholar
- Peters, J., Janzing, D., Gretton, A., & Schöölkopf, B. (2009). Kernel methods for detecting the direction of time series. In: Proccedings of the 32nd Annual Conference of the German Classification Society (GfKl 2008), 1--10.Google ScholarCross Ref
- Reichenbach, H. (1999). The direction of time. Dover.Google Scholar
- RProject (2009). The r project for statistical computing. Website, 15.1.2009, 1:07pm. http://www.r-project.org/.Google Scholar
- Shimizu, S., Hoyer, P. O., Hyvärinen, A., & Kerminen, A. (2006). A linear non-gaussian acyclic model for causal discovery. Journal of Machine Learning Research, 7, 2003--2030. Google ScholarDigital Library
- Skitovic, V. P. (1962). Linear combinations of independent random variables and the normal distribution law. Select. Transl. Math. Stat. Probab., 2, 211--228.Google Scholar
- Smola, A. J., Gretton, A., Song, L., & Schöölkopf, B. (2007). A Hilbert space embedding for distributions. In Algorithmic Learning Theory: 18th International Conference (ALT), 13--31. Springer-Verlag. Google ScholarDigital Library
- Weiss, G. (1975). Time-reversibility of linear stochastic processes. J. Appl. Prob., 12, 831--836.Google ScholarCross Ref
Index Terms
- Detecting the direction of causal time series
Recommendations
Underwater noise modeling and direction-finding based on heteroscedastic time series
We propose a new method for practical non-Gaussian and nonstationary underwater noise modeling. This model is very useful for passive sonar in shallow waters. In this application, measurement of additive noise in natural environment and exhibits shows ...
Space-time matrix method for 2-D direction-of-arrival estimation
This paper presents a novel space-time matrix method for two-dimensional direction-of-arrival estimation of multiple narrow-band sources impinging on the far field of a planar array. Using the proposed method, both the elevation angle and the azimuth ...
Direction finding using ESPRIT with interpolated arrays
The technique of interpolated arrays is applied to ESPRIT-type direction finding methods. The resulting method uses sensor arrays with an arbitrary configuration, thus eliminating the basic restrictive requirement of ESPRIT for two (or more) identical ...
Comments