No label defined (Q4258798)

Let \(X_0,X_1,\dots\) be an ergodic Markov chain with countable state space \(E=\{0,1,\dots\}\), let \(\pi=\{\pi_j\}_{j=0}^\infty\) be the stationary probability for \(X_j\), and let \(f\) be a real function on \(E\) with finite support. Denote \( S_n=\sum_{k=0}^{n-1} f(X_k) \), \( \langle\pi,f\rangle=\sum_j\pi_j f(j) \). The following statement is proved. { Theorem.} If \(\pi_j>0\) for all \(j\in E\), then \( (S_n-\langle \pi,f\rangle)/\sqrt{n} \) is asymptotically Gaussian as \(n\to\infty\). For the entire collection see [Zbl 0896.00028].