No label defined (Q4246898)

Let \(\{X_{ij}:i=1,2\); \(j=1,2,\ldots\}\) be an array of i.i.d. r.v.'s with \({\mathbb P}(X_{11}=1)={\mathbb P}(X_{11}=-1)=1/2\) and let \(S_i(n)=\sum_{j=1}^n X_{ij}.\) Let \(\xi_2(n),\) \(n=0,1,2,\ldots,\) be the number of returns to zero of \(S_2(\cdot).\) The occupation time process \(\xi^*(r,n)\) of \(R(n)=S_1(\xi_2(n))\) is defined by \[ \xi^*(r,n)=\#\big\{ k: 1\leq k\leq n,\;R(k)=r \big\},\quad r=0,\pm 1,\pm 2,\ldots . \] It is proved \[ \liminf_{n\to\infty}{\log\log n\over n} \max_r\xi^*(r,n)=\beta\quad \text{a.s.}, \] where \(\beta=0.85403\ldots.\) Previuosly \textit{E. Csáki, M. Csörgő, A. Földes} and \textit{P. Révész} [Stochastic Processes Appl. 70, No. 2, 199-217 (1997; Zbl 0911.60068)] proved that the right side does not exceed the left side. Let \[ \rho_n(0)= 0,\qquad\rho_n(l+1)=\inf\{k:k>\rho_n(l),\;S_2(k)=0\}\quad\text{if} l<\xi_2(n),\quad\rho_n\big( \xi_2(n)+1 \big)=n, \] \[ \tau(n)= \max_{0\leq i\leq\xi_2(n)}\big( \rho_n(i+1)-\rho_n(i) \big). \] For the ordered sample \(\tau_1(n)=\tau(n)\geq\tau_2(n)\geq\ldots\geq\tau_{\xi_2(n)+1}(n)\) of \(\{\rho_n(i+1)-\rho_n(i),\;i=0,1,2,\ldots,\xi_2(n)\},\) it is proved that \[ \liminf_{n\to\infty} \omega(n)\Biggl( 1-{1\over n}\sum_{i=1}^{[\omega(n)\log\log n]}\tau_i(n) \Biggr)\leq{\pi\over{2^{1/2}-1}}\quad \text{a.s.}, \] where \(\omega(n)\) is a slowly varying function with \(\omega(n)\nearrow\infty\) as \(n\to\infty,\). This is a generalization of a result previuosly obtained by \textit{E. Csáki, P. Erdős} and \textit{P. Révész} [Z. Wahrscheinlichkeitstheorie Verw. Geb. 68, 365-382 (1985; Zbl 0537.60062)]. For the entire collection see [Zbl 0901.00049].