Dynamical Analysis of Time Series by Statistical Tests
In this review we deal with the application of statistical test techniques for the extraction of structures in time series. Two kinds of questions are answered in this statistical framework: Are there any temporal dependences in the data? and Which kind of dynamics generate these temporal dependences? The first question is known as the problem of predictability and also considers the aspect of stationarity. The second question is deeper in the sense that it deals with the dynamical characterization of the detected temporal structures. Central to our approach is a cumulant-based measure of statistical dependences in Fourier space. The dynamical aspects are studied by means of the information flow. The theory is illustrated by artificial and real-world, stochastic and chaotic examples.