Alpha-diversity processes and normalized inverse-Gaussian diffusions (Q1948692): Difference between revisions
From MaRDI portal
Changed an Item |
Set profile property. |
||
Property / MaRDI profile type | |||
Property / MaRDI profile type: MaRDI publication profile / rank | |||
Normal rank |
Revision as of 05:19, 5 March 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Alpha-diversity processes and normalized inverse-Gaussian diffusions |
scientific article |
Statements
Alpha-diversity processes and normalized inverse-Gaussian diffusions (English)
0 references
24 April 2013
0 references
This paper treats processes related to the infinitely-many-neutral-alleles models. This research field is very interesting and exciting since those models have recently been extended to a class of diffusion processes associated with Gibbs partitions of two-parameter Poisson-Dirichlet type. The authors introduce a family of infinite-dimensional diffusions associated with a different subclass of Gibbs partitions, induced by normalized inverse-Gaussian random probability measures, and study natural derivation of those processes and mathematical relations between them. The first result is about a construction of a one-dimensional diffusion process, which can be seen as a dynamic version of the notion of \(\alpha\)-diversity relative to the case of normalized inverse-Gaussian random probability measures, and which plays a crucial role in the construction of processes associated with the class of normalized inverse-Gaussian random probability measures. Let \(\{ K_n(m); m \in {\mathbb N}_0 \}\) be a Markov chain with transition probabilities \[ p(k, k') = \operatorname{P} ( K_n(m+1) = k' | K_n(m) = k ) \] determined by a generalized gamma process with \(\beta \geq 0\) and \(\alpha = 1/2\), and define \(\{ \tilde{K}_n(t); t \geq 0 \}\) as \(\tilde{K}_n(t):=K_n ( [n^{3/2} t ] )/n^{\alpha}\). Let \(\{ S_t; t \geq 0 \}\) be a diffusion process driven by the stochastic differential equation \(d S_t=( \beta/ S_t) dt+\sqrt{S_t} d B_t\), \(S_t \geq 0\) with a standard Brownian motion \(B_t\). The first theorem asserts that the convergence in distribution \(\tilde{K}_n(0)\Rightarrow S_0\) leads to the convergence in distribution \[ \{ \tilde{K}_n(t); t \geq 0 \} \Longrightarrow \{ S_t ; t \geq 0 \} \text{ in } C( {\mathbb R}_+ ; {\mathbb R}_+ ) \;\text{ as } (n \to \infty). \] Next, the authors characterize such infinite-dimensional processes (like the \(\alpha\)-diversity process) in terms of their infinitesimal generator, and show that they can be obtained as the limit in distribution of a certain sequence of Feller diffusions with finitely-many types. As a matter of fact, let \({\mathcal A}\) be the operator defined by \[ \begin{aligned} {\mathcal A} &= \frac{\beta}{s} \frac{\partial}{\partial s} + \frac{1}{2} s \frac{\partial^2}{\partial s^2} + \frac{1}{2} \sum_{ i,j =1}^{\infty} z_i ( \delta_{ij} - z_j) \frac{\partial^2}{\partial z_i \partial z_j} - \frac{1}{2} \sum_{i=1}^{\infty} \left( \frac{\beta}{s} z_i + \alpha \right) \frac{\partial}{\partial z_i} \tag{1} \\ &= : {\mathcal A}_0 + {\mathcal A}_1 \end{aligned} \] with domain \({\mathcal D}({\mathcal A})\) the sub-algebra of \(C_0( {\mathbb R}_+ \times \bar{\nabla}_{\infty} )\) generated by \(f= f_0 \times f_1\); \(f_0 \in {\mathcal D}( {\mathcal A}_0 )\), \(f_1 \in {\mathcal D}( {\mathcal A}_1)\), where \[ \bar{\nabla}_{\infty} := \{ z = ( z_1, z_2, \dots ) ; z_1 \geq z_2 \geq \dotsb \geq 0, \sum_{i=1}^{\infty} z_i \leqslant 1 \}, \tag{2} \] and \({\mathcal A}_0\) (resp. \({\mathcal A}_1\)) indicates the first two terms (resp., the last two terms) in (1). Note that the closure in \(C_0 ( {\mathbb R}_+ \times \bar{\nabla}_{\infty})\) of \({\mathcal A}\) generates a strongly continuous, positive, conservative, contraction semigroup \(\{ {\mathcal T}(t) \}\) on \(C_0( {\mathbb R}_+ \times \bar{\nabla}_{\infty})\). The second main theorem asserts that, for every \(\nu\) \(\in\) \({\mathcal P}( {\mathbb R}_+ \times \bar{\nabla}_{\infty})\) (the set of Borel probability measures on \({\mathbb R}_+ \times \bar{\nabla}_{\infty}\)), there exists a strong Markov process \(Z( \cdot )\) corresponding to \(\{ {\mathcal T}(t) \}\) with initial distribution \(\nu\) and sample paths in \(C( {\mathbb R}_+; {\mathbb R}_+ \times \bar{\nabla}_{\infty})\) with probability one. Lastly, the authors discuss the association of the limit family with the class of normalized inverse-Gaussian random probability measures. As for other related works, (see, e.g., articles by \textit{S. Feng} and \textit{W. Sun} [Probab. Theory Relat. Fields 148, No. 3--4, 501--525 (2010; Zbl 1203.60120)], \textit{L. A. Petrov} [Funct. Anal. Appl. 43, No. 4, 279--296 (2009; Zbl 1204.60076); translation from Funkts. Anal. Prilozh. 43, No. 4, 45-66 (2009; Zbl 1204.60076)] and by the first two authors [Electron. Commun. Probab. 14, 501--517 (2009; Zbl 1189.60103)]).
0 references
Gibbs partitions
0 references
Poisson-Dirichlet type
0 references
generalized gamma process
0 references
infinitely-many-neutral-alleles model
0 references