Statistical Test for Dynamical Nonstationarity in Observed Time-Series Data

Abstract

Information in the time distribution of points in a state space reconstructed from observed data yields a test for ``nonstationarity''. Framed in terms of a statistical hypothesis test, this numerical algorithm can discern whether some underlying slow changes in parameters have taken place. The method examines a fundamental object in nonlinear dynamics, the geometry of orbits in state space, with corrections to overcome difficulties in real dynamical data which cause naive statistics to fail.