The reversibility condition for elementary bilinear time-series model based on output sequence alone

Abstract

Reversibility of elementary bilinear time-series model is very important issue in parametric model identification based on minimisation of mean square value of prediction error. The reason is that estimation of parameters of irreversible time-series model is irreversible it is difficult. There is already very well known mathematical reversibility condition expressed as a function of the model's coefficient, which has to be identified. The paper contains a discussion of simple condition, which can provide information about model's reversibility before its identification.