State-space and ARMA stochastic realization algorithm for nonstationary processes

A special class of nonstationary processes, namely, those generated by constant parameter models with finite starting time is considered. Using the interaction matrix for the process under consideration, a state-space realization algorithm is developed. Next a simple examination of this realization leads to the ARMA stochastic realization. Order reduction property is built into the realization algorithm. Application to reduced-order Kalman filtering is also presented.