Distinguishing random and deterministic systems: Abridged version

From MaRDI portal





The trajectories generated by some nonlinear difference equations are highly irregular and may look completely random to the naked eye. The aim of the present article is to discuss tests that are potentially useful for testing for the presence of low dimensional deterministic (nonlinear) chaos in time series data y(t) \((t=1,2,...)\). Three practical tests are discussed, which are based on the following properties (respectively): i) the correlation dimension of y(t) must be low; ii) the largest Lyapunov exponent must be positive; and iii) the residuals (or prediction errors) of an estimated (linear or nonlinear) autoregressive model fitted to y(t) must have the same correlation dimension and largest Lyapunov exponent as y(t). The test based on (iii), which is called the residual test, is used to test for deterministic chaos in U.S. real GNP, real gross private domestic investment, and Wölfer's sunspot series against the hypothesis of an AR model.




Cited in
(44)








This page was built for publication: Distinguishing random and deterministic systems: Abridged version

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1821472)