Large deviations for sums of random variables connected into a Markov chain in approximation by Poisson law (Q1382693)

Let \(({\mathcal X}, {\mathcal F})\) be a measurable space and let \(P=P(x,A)\) be a probability transition function on it. Consider a stationary Markov chain \(\xi_i\), \(i=0,1,2,\ldots\), on the state space \(({\mathcal X},{\mathcal F})\) with transition probabilities in one step \(P\{\xi_i\in A\mid\xi_{i-1}=x\}=P(x,A)\), \(i=1,2,\ldots\). Let, for every \(n=1,2,\ldots\), \(f_n(x)\) be a measurable function on \(({\mathcal X},{\mathcal F})\) taking non-negative integer values, \(S_n:= f(\xi_1)+f(\xi_2)+\cdots+f(\xi_n)\), \(b_n:=ES_n=nf(\xi_1)\). Denote \(\Pi(x,b):=\sum_{k<x}b^k e^b /k!\) (the Poisson distribution with parameter \(b\)). In case when \(P(x,A)\) satisfies the condition of uniformly strong mixing, the limit theorem for large deviations holds: under some specific conditions \[ {{P\{S_n\geq x_n\}\over{1-\Pi(x_n,b_n)}}}-1=O\bigg( {{x_n^2}\over{\Delta_n}}\bigg) \] when \(x_n=o(\Delta_n^{1/2})\). Exponential estimates for probabilities of large deviations \(P(S_n\geq x)\) can also be obtained. The aim of this paper is to prove more general results when Markov chains are \(L^p\)-regular [|see the author, Lith. Math. J. 22, 265-275 (1983); translation from Litov. Mat. Sb. 22, No. 3, 67-80 (1982; Zbl 0531.60064)].