Answers are grounded in this paper and cite their sources.
This paper came out of an attempt to solve the following general problem: Suppose ₙ, n 1\ is a stationary process with a finite state-space J. Under what conditions can we express it as a function of a finite Markov chain? More precisely, when can we find a stationary Markov chain ₙ, n 1\ with a finite state-space I and a function f on I onto J such that the process (Xₙ)\ has the same distribution as ₙ\? We do not solve the general problem here but for mixing processes we obtain a theorem which is the best possible in a certain sense. Suppose ε denotes a state of J and suppose $s, t$ denote finite sequences of states of J. If s = ε₁ ⋯ εₙ, let p(s) = P (Y₁, ⋯, Yₙ) = s. For each ε, define n(ε) to be the largest integer n such that we can find sᵢ, tᵢ (i = 1, ⋯, n) such that the matrix \|p(sᵢε tⱼ)\| is nonsingular. Gilbert [4] has shown that if ₙ\ is a function f of a finite Markov chain ₙ\ and if f takes N(ε) states of I into the state ε of J, then n(ε) N(ε). If n(ε) = N(ε), then ₙ\ is said to be a regular function of a Markov chain. Thus a necessary condition for ₙ\ to be a function of a finite Markov chain is that ∑ n(ε) is finite. It is proved here that if ∑ n(ε) < ∞ and if the process ₙ\ is mixing, then there exists a positive integer m^ such that for every m m^ the process + 1, n 0\ is a function of a Markov chain with ∑ n(ε) states. An example is constructed to show that m^ cannot, in general, be brought down to 1. Thus the whole process ₙ, n 1\ may still not be a function of a Markov chain with ∑ n(ε) states.
No takes yet. Share an insight, caveat, or question.
S. W. Dharmadhikari (1963) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: