On Random Fragmentations Arising From Binary Splitting

From MaRDI portal





As discussed by \textit{T. Huillet} and \textit{B. Jeannet} [J. Phys. I, 6, 245--256 (1996)] for the similar but deterministic construction, the authors' main interest of such cascading models consists in their capability to account for a coexistence of phase phenomenon; that is to say, in those models it is observed that iterative subdivision can come randomly to a halt at any node and microscopic and macroscopic fragments are likely to coexist in the limiting fragmentation tree. The purpose of this paper is to discuss some statistical problems arising in partitioning space according to such a structure as seen in the random fragmentation model with binary splitting. More precisely, the authors consider a homogeneous fragmentation model for randomly fragmenting an interval based on random conservative binary cascades, and such a model can be understood in terms of binary Galton-Watson (BGW) branching process, at each step of which there are a probability \(p\) for splitting and a probability \(1-p\) for not splitting. When splitting occurs, parental mass is shared between the two child fragments at random, and the partition law is a Dirichlet distribution \(D_2(\theta)\) with \(\theta >0\). Since it is a two-parameter model, the symbol BGW\((\theta, p)\) is used. Some statistical features of this BGW\((\theta, p)\)-fragmentation model are studied, such as the fragment size distribution, the partition and moment functions of mass, Rényi's average fragment size, the sizes of smallest and largest fragments, and the contribution to total mass of small fragments. This model exhibits a phase transition between sub-critical and super-critical regimes, which is discussed as well. Some problems related to size-biased or unbiased sampling from BGW\((\theta , p)\) model are also investigated. In particular, the principal topics are as follows: the size of a size-biased sample, the height of the most recent common ancester of two distinct size-biased sample fragments, and the number of distinct observations in a \(k\)-sample from BGW\((\theta, p)\). For some other related works, see e.g. \textit{J. Barrera, T. Huillet} and \textit{C. Paroissin} [Probab. Eng. Inf. Sci. 19, No. 1, 83--97 (2005; Zbl 1077.60005)]. As for additional physical motivations of cascading models, see e.g. \textit{P. L. Krapivsky} and \textit{I. Grosse} [Phys. Rev. E 61, R993--R996 (2000)] and \textit{P. L. Krapivsky, E. Ben-Naim} and \textit{I. Grosse} [J. Phys. A, Math. Gen. 37, No. 8, 2863--2880 (2004; Zbl 1038.82028)]. In addition, as to some other models with similar construction from the theory of percolation and contact processes, see e.g. \textit{T. M. Liggett} [Ann. Appl. Probab. 5, No. 3, 613--636 (1995; Zbl 0842.60090)].











This page was built for publication: On Random Fragmentations Arising From Binary Splitting

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5711160)