Modelling and analysis of non-linear time series

Abstract

Abstract The modelling of non-linear time series is reviewed and new results are introduced by employing some ideas from the identification of non-linear control systems. Both global and local conditions for stationary and invertibility are established for the general non-linear time-series model and it is shown how these results provide a framework for time-series estimators. Methods of computing multistep-ahead predictions are studied and the usefulness of polynomial models is discussed.