Invariance principles for renewal processes (Q1097572): Difference between revisions

Let Z(t), \(t\geq 0\), be a real-valued stochastic process and let \(N(t)=\inf \{x:Z(x)>t\}\), \(t\geq 0\), be its inverse, or renewal process. The aim of the paper is to construct best approximations for N(t). Two main theorems are proved. The first of them is as follows: If r(T)\(\uparrow \infty\) and r(T)/T\(\downarrow 0\) as \(T\to \infty\) and for some standard Wiener process W(t) holds \[ A_ T=\sup_{0\leq t\leq T}| (Z(t)-\mu t)/\sigma -W(t)| =O(r(T))\;a.s., \] then there exists another standard Wiener process \(\hat W(t)\) such that \[ B_ T=\sup_{0\leq t\leq T}| (N(t)-t/\mu)\mu^{3/2}/\sigma -\hat W(t)| =O(\log T+\quad r(t))\quad a.s. \] The second theorem says that if some probability inequality holds for \(A_ T\), then a similar inequality holds also for \(B_ T\). These two results provide a general method for proving invariance principles for renewal processes. Several applications are given. Among them the i.i.d. case is studied in details. In this case the obtained rates are the best possible.