Long random walk excursions and local time (Q1198592)

The authors treat some technical properties of symmetric random walks relating its excursions with local time. Thus let \(\{X_ n,\;n\geq 1\}\) be a sequence of symmetric i.i.d. Bernoulli random variables and \(S_ n\) be its partial sum (or random walk) process, \(S_ 0=0\). Define its characteristics \[ \xi(x,n)=\#(k: 0<k\leq n, S_ k=n),\;n\in\mathbb{N},\;x\in\mathbb{Z}; \] \[ p_ 1(x)=\inf\{n: n>0, S_ n=x\};\;\text{and recursively }\rho_{i+1}(x)=\inf\{n: n>\rho_ i(x), S_ n=x\}. \] Finally let \(F\) be the distribution function of \(\rho_ 1\) and \(M(a,x,n)\) be the number of excursions of duration greater than \(a\) of \(S_ k\), \(k\geq 0\), away from \(x\) and are completed by time \(n\). Theorem. Let \(0\leq a_ n\uparrow\), \(a_ n\in\mathbb{N}\), \(a_ n\leq n(\log n)^{-\beta}\) for \(\beta>2\). Then for each \(x\in\mathbb{Z}\), one has (i) \(P[\lim_{n\to\infty}(a_ n/n)^{1/4}(\log\log n)^{-1}| M(a_ n,x,n)-(1-F(a_ n))\xi(x,n)|=0]=1\), (ii) if \(a=n(\log n)^{-\beta}\), \(0<\beta\leq 2\), then \(P[M(a_ n,x,n)=0\) infinitely often]=0. Another result of a similar kind is also given.