A remarkable \(\sigma\)-finite measure unifying supremum penalisations for a stable Lévy process (Q376688): Difference between revisions
From MaRDI portal
Latest revision as of 10:43, 30 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A remarkable \(\sigma\)-finite measure unifying supremum penalisations for a stable Lévy process |
scientific article |
Statements
A remarkable \(\sigma\)-finite measure unifying supremum penalisations for a stable Lévy process (English)
0 references
19 November 2013
0 references
\textit{B. Roynette, P. Vallois} and \textit{M. Yor} considered penalisations of the Brownian motion [Stud. Sci. Math. Hung. 43, No. 2, 171--246 (2006; Zbl 1121.60027); ibid. 43, No. 3, 295--360 (2006; Zbl 1121.60004); Jpn. J. Math. (3) 1, No. 1, 263--290 (2006; Zbl 1160.60315)]: Let \(X=((X_t, \mathcal{F}_t)_{t\geq 0}, W)\) be the canonical representation of the one-dimensional Brownian motion with supremum process \(S = (S_t =\sup_{s\leq t} X_s)_{0\leq t\leq \infty}\) and let \(f\) be a density, \(f\geq 0\), \(\int f d x=1\). Then there exists a unique probability measure \(W^{(f)}\) on \(\mathcal{F}_\infty = \sigma(\bigvee \mathcal{F}_t)\) such that for any time \(s>0\) and any bounded \(\mathcal{F}_s\)-measurable functional \(F_s\) we have \(W[f(S_t)F_s]/W[f(S_s)]\to W^{(f)}[F_s]\) (with \(t\to\infty\)), and \(W^{(f)}|_{\mathcal{F}_s} = M_s^{(f)} W|_{\mathcal{F}_s}\), for the Azema-Yor martingale \[ \left(\left(M_s^{(f)} = f(S_s)(S_s-X_s) + \int_{S_s}^\infty f(x) d x\right)_{s\geq 0}, W\right). \] In addition, \(W^{(f)}(S_\infty\in d x) = f(x)d x\) and the process \(X\) is decomposed into independent processes \((X_u: u\leq g)\) and \((X_g-X_u: u\geq 0)\), where \(g\) denotes the first hitting time of the supremum. This result and related investigations are partially generalized into various directions; e.g., replacing the supremum process by other kinds of weighted processes, in particular, by `local times', resp. the `Kac killing penalisation'. Furthermore, the Brownian motion is replaced by other Lévy processes satisfying additional conditions (fulfilled in the case of Brownian motion), in particular, for stable processes. (Of course, there the Wiener measure \(W\) has to be replaced by a (suitable) probability \(\operatorname{P}\) and \(W^{(f)}\) by a probability \(\operatorname{P}^{(f)}\)). In the main part, the author introduces a new \(\sigma\)-finite measure \(\operatorname{P}_{\mathrm{sup}}\) which unifies previous `supremum-penalisations'. In particular, the absolute continuity relation \(f(S_\infty)\cdot \operatorname{P}_{\mathrm{sup}}/ \operatorname{P}_{\mathrm{sup}}[f(S_\infty)] \) \(= \operatorname{P}^{(f)}\) on \(\mathcal{F}_\infty\) is proved. The paper is organized as follows: The first sections describe the state of previous investigations, including a survey of properties of Lévy processes and, in particular, stable processes as well as path decompositions of a Lévy process into independent pre- and post-supremum processes (i.e., as afore mentioned for the Brownian motion, decompositions of the process at the time when the supremum is attained). In Section 6, the author defines the measures \(\operatorname{P}^{(f)}\) and generalized Azema-Yor martingales. The afore mentioned measure \(\operatorname{P}_{\mathrm{sup}}\), unifying the supremum-penalisations, and its properties appear in Section 7. Reviewer's remark: It should be mentioned that `penalisations of Lévy processes' is an active field of investigations up to present; see, e.g. [\textit{B. Roynette, P. Valois} and \textit{M. Yor}, ESAIM, Probab. Stat. 13, 152-180 (2009; Zbl 1189.60069), and \textit{B. Roynette} and \textit{M. Yor}, ibid. 14, 65--92 (2010; Zbl 1219.60036)], and the literature mentioned there.
0 references
Lévy process
0 references
stable process
0 references
reflected process
0 references
penalisation
0 references
Chaumont's \(h\)-transform process
0 references
path decomposition
0 references
conditioning to stay positive/negative
0 references
conditioning to hit \(0\) continuously
0 references
0 references
0 references
0 references
0 references
0 references