On the nonlinear renewal theorem (Q2638675)

Let \(Z_ 1,Z_ 2,..\). be a sequence of random variables that has infinite supremum with probability one. Let \(t_ a\) be the first n with \(Z_ n\geq a\) and let \(H_ a\) be the distribution function of the excess \(Z_{t_ a}-a\) over the boundary a. Interest is in conditions which ensure (weak) convergence of \(H_ a\) as \(a\to \infty\); a process with this property is called a renewal process. The classical case on which the generalizations are modelled has \(\{Z_ k\}\) a random walk with positive drift. The present paper gives conditions for weak convergence of \(H_ a\) which are too intricate to be repeated here in detail; they include the condition that the joint conditional distribution of the first m increments following \(t_ a\), given the history \({\mathcal F}_ a\) of the process up to \(t_ a\), converges to the distribution of the initial m-segment of an appropriate conditional renewal process. The conditioning random variable for the latter must arise as the limit in distribution of \({\mathcal F}_ a\)-measurable random variables. The result is applied to the Euclidean norm of multidimensional driftless random walks.