Asymptotic Theory of Large Deviations for Markov Chains

From MaRDI portal





Let \(p_n(x,y)\) denote the density function of a real-valued stationary ergodic Markov chain \(X_n\) and its sample averages \(Y_n:={1\over n}\sum_{k=1}^n X_k\), i.e., \(p_n(x,y) dx dy=P(X_n\in dx, Y_n\in dy)\). This paper deals with the formal asymptotic expansion \[ p_n(x,y)=e^{-n\psi(n)}\sqrt{n} \Big[q^0(x,y)+ {q^1(x,y)\over n}+{q^2(x,y)\over n^2}+\cdots\Big]. \] Here, \(\psi\) is the classical large deviation rate function. The coefficients \(q^0(x,y)\) and \(q^1(x,y)\) are calculated explicitly, and an algorithm is proposed in order to deal with the higher order terms.











This page was built for publication: Asymptotic Theory of Large Deviations for Markov Chains

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4210068)