Important issues of system identification using algorithms for structure selection: do noise terms have any influence when the data are chaotic?

Addresses one of the problems of model structure selection of nonlinear systems behaving chaotically. It is shown that the noise terms can determine whether an estimated model is dynamically valid or not, even if the process terms of such a model are perfectly correct.