The first passage time of a stable process conditioned to not overshoot (Q325892)

Let \((X_t)_{t\geq0}\) be a strictly \(\alpha\)-stable Lévy process for \(\alpha\in(0,2)\) with positivity parameter \(\rho=P\{X_1>0\}>0\) and let \(T_x=\inf\{s>0: X_s\geq x\}\) denote its first-passage time above the level \(x>0\). The author first gives an alternative and constructive proof of the known fact that the distribution of \(T_x\) is absolutely continuous with respect to the Lebesgue measure. An important role is played by the distributional limit of \(T_x\) conditioned that the overshoot does not extend \(\varepsilon\) as \(\varepsilon\downarrow0\) (obtained by the author in [Theory Probab. Appl. 55, No. 4, 683--729 (2011; Zbl 1238.60052)] and denoted by \(T_x^0\)). As a consequence, the density function of \(T_x\) is given in terms of the distribution of \(T_x^0\). Further, a relation between \(T_x^0\) and the dual process conditioned to die at 0 is presented. This allows to relate the density of \(T_x\) to other classical results in fluctuation theory of stable Lévy processes. In particular, relations to representations of the density of \(\sup\{X_t:t\in[0,1]\}\), obtained by \textit{R. A. Doney} and \textit{M. S. Savov} [Ann. Probab. 38, No. 1, 316--326 (2010; Zbl 1185.60052)] and \textit{L. Chaumont} [Ann. Probab. 41, No. 3A, 1191--1217 (2013; Zbl 1277.60081)], are given, which allow the author to represent the densities of \(T_x\) and \(T_x^0\) in terms of the densities of entrance laws of the reflected excursions at the minimum or the densities of terminal values of the dual stable meander.