Invariance principles for local times at the maximum of random walks and Lévy processes (Q989179)

For a sequence of Lévy processes \(X^{(n)}\) the authors study the convergence of the sequence \(L^{(n)}\) of local times at the supremum of \(X^{(n)}\) to the local time at the supremum of the limiting Lévy process \(X\) (note that a sequence of local times at the fixed level of \(X^{(n)}\) does not necessarily converge to the local time of \(X\)). The analogous problem is considered also for random walks \(S^{(n)}=(S^{(n)}_k, k=0,1,\dots)\), \(n\in \mathbb{N}\). The ladder time and ladder height process are introduced for \(S^{(n)}\). It is proved that whenever \((S^{(n)}_{[nt]},t\geq 0)\) converges in law in the Skorokhod space \(D([0,\infty))\) to \(X\) (such that \(0\) is regular for the half-line \((0,\infty)\)), a normed version of the bivariate ladder process of \(S^{(n)}\) converges in law to the (ascending) bivariate ladder process of \(X\). This result extends Theorem 3.2 by \textit{P. Greenwood, E. Omey} and \textit{J. L. Teugels} [Z. Wahrsch. Verw. Geb. 61, 527--539 (1982; Zbl 0493.60072)]. Moreover, if \((S^{(n)}_{[nt]}, t\geq 0)\to (X_t,t\geq 0)\) a.s., \(n\to \infty\), then a normed version of the local time at the maximum of \(S^{(n)}\) converges in probability uniformly on compact subsets to the local time at the supremum of \(X\). The similar result is established for a sequence \(X^{(n)}\) of Lévy processes converging a.s. to \(X\). Various applications are given to conditional processes. For instance, an invariance principle for the mender (extending known results for the case of attraction to a stable law) is obtained.