Forwards, backwards, and dynamically reversible Markovian models of second-order processes

Related causal and anticausal models of second-order processes are studied; new properties are characterized in terms of associated matrix fraction descriptions. Situations in which related causal and anticausal models have the same transfer function matrix are examined in detail, and the models are shown to possess internal and external time-reversiblity properties. Connections with ideas of statistical mechanics and passive networks are also indicated.