scientific article; zbMATH DE number 4028580

From MaRDI portal
Publication:3769741





Let \(x_ t\) be a semi-Markov process with finite state space I. Let us introduce the notations: \[ \tau:=\inf \{t>0;\quad x_ t\neq x_ 0\},\quad F_{ij}(t):=P_ i\{\tau \leq t;\quad x_{\tau}=j\}, \] \[ F_ i(t):=\sum_{j\in I}F_{ij}(t)=P_ i(\tau \leq t),\quad \phi_ i(s):=\int^{\infty}_{0}e^{-st}dF_ i(t),\quad i,j\in I, \] where \(P_ i\) is the conditional probability under the condition \(x_ 0=i\). The matrix \((p_{ij})_{i,j\in I}\) \((p_{ij}:=P_ i\{x_{\tau}=j\})\) of the embedded Markov chain is supposed to be irreducible, and its stationary distribution will be denoted by \(\{p_ j\}_{j\in I}\). If g is a measurable function from \({\mathbb{R}}^ 2\) to [0,\(\infty)\) then \(g(x_ t,t)\) is a process with Markovian interference time. The result of the paper is: If there exist a real number \(\alpha\in [0,1)\) and a (slowly varying at zero) function L such that \[ [1-\phi_ i(s)]/s^{\alpha}L(s)\to a_ i(s\to 0),\quad i\in I,\quad \sum_{i\in I}a_ ip_ i>0, \] \[ t^{-1}\int^{t}_{0}f(g(j,u))du\to b_ j\quad (t\to \infty), \] where f is a bounded, positive measurable function on [0,\(\infty)\), then \[ P_ i\{t^{- 1}\int^{t}_{0}f(g(x_ u,u))du<x\}\to G(x),\quad t\to \infty,\quad where \] \[ \int [dG(x)/(\lambda +x)]=\sum_{i\in I}a_ ip_ i(b_ i+\lambda)^{-1}/\sum_{i\in I}a_ ip_ i(b_ i+\lambda)^{\alpha}. \] A previous limit theorem by the same author is obtained when taking the special case \(g(i,u):=i\).











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3769741)