A limitation of Markov representation for stationary processes (Q797215)

In this paper the authors consider the problem of representing a strictly stationary process \((\xi_ n\); \(n\in {\mathbb{Z}})\) by a real-valued, stationary Markov chain \((Y_ n\); \(n\in {\mathbb{Z}})\) in the form \(\xi_ n=f(Y_ n)\), where f is a measurable function. The authors show that the existence of such a representation imposes important restrictions on the distribution of the process. They construct examples of countably valued stationary processes satisfying strong mixing properties such that this representation does not exist.