Normalized excursion, meander and bridge for stable Lévy processes (Q1365486)

The author introduces the notions of normalized excursion, meander, and bridge for \(\alpha\)-stable Lévy processes \(\{X_t\}_{t\geq 0}\) (filtration \(\{{\mathfrak F}_t\}_{t\geq 0})\) and the reflected process \(\{X_t- \inf_{s\leq t} X_s\}_{t\geq 0}\). If \(\underline n(\bullet)\) is a measure on the path space such that \(\underline n(\zeta >t) =\Gamma (\rho)^{-1} t^{\rho-t}\) \((\zeta\) is the life-time, \(\rho: =\mathbb{P}^0 (X_1\geq 0))\), the normalized excursion law (of the reflected process) is given by \[ \int F(\omega_\bullet) \mathbb{P}^{(e)} (d\omega) =\underline n \biggl( F\bigl( \zeta (\omega)^{-1/ \alpha} \omega_{\zeta (\omega)_\bullet} \bigr) \mid \zeta (\omega)>t \biggr)/ \underline n(\zeta >t). \] \(F\) denotes a mesurable functional on the path space. The meander of length 1 is given by \[ \int F(\omega_\bullet) \mathbb{P}^{(m)} (d\omega) =\underline n \biggl(F \bigl(k_1 (\omega) \bigr) \mid \zeta (\omega)> 1\biggr) \] where \(k_1\) is the usual killing operator. One can see that \(\mathbb{P}^{(m)}\) is the law of the reflected process \(\{X_s- \inf_{r\leq s} X_r\}_{g_t \leq s\leq t}\) under the restriction that \(t-g_t =1\) and \(g_t\) stands for the last zero of the reflected process before time \(t\). The process \(\{X_t\}_{t\geq 0}\) conditioned to stay positive is defined via an \(h\)-transform, \(h(x)= \text{const} x^{\alpha (1-\rho)}\) and its law \(\mathbb{P}^\uparrow\) is given by \[ \mathbb{P}^\uparrow (\Lambda) =\underline n\bigl( h(X_t) \mathbf{1}_{\{ \Lambda; \zeta>1\}} \bigr), \quad \Lambda \in {\mathfrak F}_t. \] In the first part of the paper it is shown that \(\mathbb{P}^{(e)}\) and \(\mathbb{P}^{(m)}\) are absolutely continuous and that \(\mathbb{P}^{(e)}(\bullet\mid X_{1-}=x)=\mathbb{P}^{(m)}(\bullet \mid X_{1-} =x)\). Moreover, \(\mathbb{P}^{(m)} (\bullet)=\mathbb{E}^\uparrow(\Gamma(\rho) h(X_1)^{-1}\mathbf{1}_\bullet)\) and one can even construct the meander from the original process (conditioned to stay positive) via \[ \int F(\omega_\bullet)\mathbb{P}^{(m)} (d\omega) =\lim_{\varepsilon \to 0} \mathbb{E}^\uparrow \bigl(F (\omega_\bullet) \mid X_1-\underline{\underline X}_1\leq\varepsilon\bigr) \] where \(\underline{\underline X}_1 =\inf_{s \geq 1} X_s\). In the second part the author investigates the relation of the (normal) bridge process \((\{X_t\}_{t\geq 0}, \mathbb{P}^1_{0,0})\) of length 1 and \(X_0= X_1=0\) and the bridge process \((\{X_t \}_{t\geq 0},\mathbb{P}^{\uparrow,1}_{0,0})\) under \(\mathbb{P}^\uparrow\) that is, roughly, the process \(\{X_t \}_{t\geq 0}\) conditioned to stay positive and to be zero at time 1. More precisely, \[ \mathbb{P}^{\uparrow,1}_{0,0} (\Lambda)=\mathbb{E}^\uparrow\left(\mathbf{1}_\Lambda c_t {j^*_{t-s} (X_s) \over h(X_s)} \right), \quad \Lambda \in {\mathfrak F}_s,\;s\leq t, \] with \(h\) as above, some constant \(c_t\), and the entrance law \(j^*_t\) under \(\underline n\) of the dual process \((\{-X_t\}_{t\geq 0},\mathbb{P})\). If \(m\) denotes the unique instant, the normal bridge \((\{X_t\}_{t\geq 0}, \mathbb{P}^1_{0,0})\) reaches its minimum, and if \(U\) is some uniformly distributed (under \(\mathbb{P}^{0,\uparrow}_{0,0})\) random variable independent of \((\{X_t\}_{t\geq 0},\mathbb{P}^{\uparrow, 1}_{0,0})\), then the relation of \(\mathbb{P}^1_{0,0}\) and \(\mathbb{P}^{\uparrow,1}_{0,0}\) is given by: (1) \(m\) and \(\{X_{m+t \pmod 1} -X_m \}_{0\leq t\leq 1}\) are independent w.r.t. \(\mathbb{P}^{\uparrow,1}_{0,0}\) and, under \(\mathbb{P}^1_{0,0}\), \(m\) is uniformly distributed and \(\{X_{m+t \pmod 1} -X_m\}_{0\leq t\leq 1}\) follows \(\mathbb{P}^{\uparrow,1}_{0,0}\). (2) Under \(\mathbb{P}^{\uparrow,1}_{0,0}\), the process \(\{X_{U+ t \pmod 1} -X_U\}_{0\leq t\leq 1}\) has the law \(\mathbb{P}^1_{0,0}\).