Dependence Identification in a Time Series on the Basis of Structural Difference Schemes

摘要

The method of dependence identification is described, in which each model is compared to a linear or nonlinear structural difference scheme. Inclusion of nonlinear difference schemes into structural models significantly expands the number of identifiable dependences. This method makes it possible to choose the sought model among the given set of dependences. The model chosen is a model for which the distance between the vector of estimates of autoregression coefficient and the corresponding tolerance range of coefficients of the structural difference scheme is minimum. This method was validated via statistical modeling by the Monte Carlo method.