Answers are grounded in this paper and cite their sources.
A stationary sequence ₙ:n = 1, 2, ⋯\ of random variables with D values (states) is said to be a function of a finite Markov chain if there is an integer N D, an N × N irreducible aperiodic Markov matrix M, a stationary Markov chain ₙ\ with transition matrix M, and a function f such that Yₙ = f(Xₙ). For any finite sequence s of states of ₙ\, let p(s) = P\(Y₁, ⋯, Yₙ) = s\. For any state ε, let sε t be the sequence s followed by ε followed by the sequence t. For every state ε, let n(ε) be the largest integer n such that there are finite sequences s₁, ⋯, sₙ, t₁, ⋯, tₙ such that the matrix \| p(sᵢε tⱼ):1 i, j n \| is nonsingular. If ₙ\ is a function of a finite Markov chain, then Σ n(ε) N. There is a finite set ₁, ⋯, sN, t₁, ⋯, tN\ of finite sequences such that $p(s)$ satisfies the recurrence relations {equation*}{1}p(sε t) = ∑f(i)=ε a_i(s)p(s_iε t),{equation*} where aᵢ(s) either is zero for all s or else is a ratio of determinants involving only p(sε tₖ) and p(sⱼε tₖ) for $f(j) = f(k) = f(i)$. If ₙ\ has D states and is a function of a Markov chain having N states, then the entire distribution of ₙ\ is determined by the distribution of sequences of length 2(N - D + 1). For each N and D, a function of a Markov chain is exhibited which attains this bound. If there is a Markov chain ₙ\ with N = Σ n(ε) states such that ₙ\ is a function of ₙ\, then ₙ\ is said to be a regular function of a Markov chain. If ₙ\ is a regular function of a Markov chain having transition matrix M, then M = X⁻¹ AX, where A is an N × N matrix with elements aᵢⱼ = aⱼ(sᵢf(i))--defined by (1) above. X = \|xᵢⱼ\| is a nonsingular N × N matrix such that xᵢⱼ = 0 unless $f(i) = f(j)$, the first row of each nonzero submatrix along the diagonal consists of positive numbers, and Σⱼxᵢⱼ = p(sᵢf(i)). Any N × N Markov matrix giving the same distribution for ₙ\ can be written in this form, with the same A and with an X having the above properties. Any matrix of this form which has all elements nonnegative is a Markov matrix giving the same distribution for ₙ\. There are Σ(ε)\² - N "unidentifiable" parameters in the matrix X, and at most N² - Σ(ε)\² "identifiable" parameters, determined by the distribution of ₙ\, in the matrix A.
No takes yet. Share an insight, caveat, or question.
Edgar J. Gilbert (1959) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: