CHAOS VERSUS NOISE IN EXPERIMENTAL DATA

Abstract

Problems of the distinction between stochastic and deterministically chaotic data are discussed. Conventional attempts such as calculation of dimensions, entropies, or Lyapunov exponents are shown to be prone to errors, or outright useless. We test the recently proposed method of surrogate data with time series from numerical and from real world experiments. This technique turns out to be a useful tool for the distinction, far superior to previous approaches.