Modèles paramétriques de processus de branchement uni et multi-types

Abstract : This thesis aims to propose parametric models for single and multi-type branching processes. The importance of the theory of branching processes is pointed out. Hence, developing various tools and specific concepts in several domains is important for applications. For those purpose, we recall some definitions and results of the single-and-multi-type branching processes theory in discrete and continuous case. Afterward, we focus on the methodological development of those models.In the second part, the evolution of a single population in the continuous case has been studied. Then, some parametric distribution families associated to particular branching mechanisms are explored. Recursive computational procedure and relevant properties concerning the associted probability distributions are derived from generating functions that satisfy specified linear partial differential equations. The suggested families are useful for the modeling of systems that are more coherent with population dynamics, contrarily to the usual hypothesis of Poisson distributions, that cannot be argued.In the third part, the evolution of different populations with interaction is explored. Similarly, some parametric models of homogeneous multi-type branching processes in continuous time are proposed. Afterwards, we consider a particular model where an autonomous donor parent population feeds in individuals, K types progeny populations that interacts. This model is well adapted to the study of dynamical systems of populations in interaction. This simple model, but has a rich variety of behaviors.The study of such systems is also done regarding the evolution of generating functions of multidimensional ndividual countrings. To achievea such study, ordinary and partial differential equations are used to establish the implicit equations of temporal and multidimensional distributions. Analytical and numerical methods for equation resolution are then discussed, and examples of particular models are developed.In conclusion, the relevancy of this approach is argumed, censidering parameters interpretation in the development of inference methods for the various applied domains.
Document type :
Theses
Complete list of metadatas

Cited literature [72 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02160476
Contributor : Abes Star <>
Submitted on : Wednesday, June 19, 2019 - 5:06:19 PM
Last modification on : Thursday, June 20, 2019 - 1:26:32 AM

File

OUAARI_2018_archivage_cor.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02160476, version 1

Collections

Citation

Amel Ouaari. Modèles paramétriques de processus de branchement uni et multi-types. Statistiques [math.ST]. Université Montpellier, 2018. Français. ⟨NNT : 2018MONTS109⟩. ⟨tel-02160476⟩

Share

Metrics

Record views

21

Files downloads

28