Correlation integral as a tool for distinguishing between dynamics and statistics in time series data (Q1964246)

The authors propose a simple criterion for time series analysis which is based on the correlation integral. They consider the following classes of data: (a) data from smooth low-dimensional dynamical systems, (b) data from high-dimensional systems, (c) data from deterministic but nonsmooth systems, (d) stationary random processes, (e) nonstationary processes. Note that in most cases the correlation integral is a good tool, since its calculation usually requires less data points than many others.