Trimmed Lévy processes and their extremal components (Q1986022)

The article investigates the joint weak limiting behaviour of \((^{(r)}X_t, \Delta_t^{(r)})\) for fixed \(t\) as \(r \rightarrow \infty.\) Here, \(\Delta_t^{(r)}\) is the \(r\)th largest jump of \(X_t\) on \([0,t], t > 0, r = 1, 2, \ldots,\) and \(^{(r)}X_t\) is the \textit{trimmed subordinator} defined as \(^{(r)}X_t = X_t - \sum_{i=1}^{r}\Delta_t^{(i)},\) \((X_t)_{t \geq 0}\) being a driftless subordinator with jump process \((\Delta_t:= X_t - X_{t-})_{t > 0},\) having infinite Lévy measure \(\Pi\) with tail function \(\overline{\Pi}(x):= \Pi\{(x, \infty)\}, x > 0,\) satisfying \(\overline{\Pi}(0+) = \lim_{x \downarrow 0} \overline{\Pi}(x) = \infty,\) (``infinite activity''). This is done by first studying the convergence of \(\Delta_t^{(r)}\) as \(r \rightarrow \infty,\) for fixed \(t > 0,\) assuming conditions on \(\overline{\Pi}\) guaranteering that \(\Delta_t^{(r)}\) has a limit distribution under linear deterministic normalization, and then proving that a normal limit distribution exists under a random centering and norming for \(^{(r)}X_t,\) conditional on the value of \(\Delta_t^{(r)}.\) The article then investigates replacing the random centering and norming with deterministic versions after extending the asymptotic normality of \(^{(r)}X_t,\) conditional on the value of \(\Delta_t^{(r)},\) to a joint weak limit for \((^{(r)}X_t, \Delta_t^{(r)})\) in which the limit has independent components, under random centering and norming. An example for a stable driftless subordinator is discussed for a particular case.