Techniques for bilinear time series analysis

This paper reviews the general problem of nonlinear time series analysis. The special case of bilinear time series analysis is discussed in detail. The stability of the estimated nonlinear system models is of particular importance. We discuss a simple sufficient condition for the stability of such models. Several approaches to the estimation of the parameters of a bilinear time series are then reviewed. The paper concludes with a discussion of some theoretical aspects associated with using nonlinear time series models in applications involving predictive data compression.<
>