Smoothness, degrees of freedom and Liapunov exponents of a time series

Abstract

We propose a set of tests addressing the issue of determining whether the generating law of a time series is a stochastic process or a chaotic dynamics. In the latter case, we test the smoothness and find the number of degrees of freedom of the underlying dynamics. We propose an adaptation of Eckmann and Ruelle algorithm for the computation of the Liapunov exponents of a time series. This algorithm computes efficiently the whole Liapunov spectrum of the observed dynamics, avoiding the problem of the spurious exponents.