On the asymptotics of distributions in two-sided boundary problems for random walks defined on a Markov chain (Q809479)

[For the entire collection see Zbl 0697.00027.] The homogeneous two-dimensional Markov chain \((S_ n,\kappa_ n)^{\infty}_ 0\) with the state space \(R\times \{1,...,m\}\) (m\(\geq 1)\) is considered. The following conditions are assumed: \[ P(S_ n<t,\kappa_ n=k | S_{n-1}=s,\kappa_{n-1}=j)=P(S_ n<t- s,\kappa_ n=k | S_{n-1}=0,\kappa_{n-1}=j) \] (s,t\(\in R\); j,k\(\in \{1,...,m\}\); \(n\geq 1)\) and \((\kappa_ n)^{\infty}_ 0\) is an indecomposable and aperiodic Markov chain. The author obtains full asymptotic expansions in powers of 1/\(\sqrt{n}\), a/n, b/n for conditional probabilities \(P(S_ N\in A,N=n,\kappa_ n=k | \kappa_ 0=j)\) and \(P(S_ n\in B,N>n,\kappa_ n=k | \kappa_ 0=j),\) where \(N=\min \{n:\) \(S_ n\not\in (-a,b)\}\) \((a,b>0)\); A, B are some intervals. He essentially uses methods of the complex analysis which are applied to inversion of generating functions (on the variable n) of these probabilities.