Search

We introduce a class of branching processes in which the reproduction or lifetime distribution at a given time depends on the total cumulative number of individuals who have been born in the population until that time. We focus on a continuous-time version of these processes, called total-progeny-dependent birth-and-death processes, and study some of their properties through the analysis of their fluid (deterministic) approximation. These properties include the maximum population size, the total progeny size at extinction, the time to reach the maximum population size, and the time until extinction. As the fluid approach does not allow us to approximate the time until extinction directly, we propose several methods to complement this approach. We also use the fluid approach to study the behaviour of the processes as we increase the magnitude of the individual birth rate.

We consider discrete-time parametric population-size-dependent branching processes (PSDBPs) with almost sure extinction and propose a new class of weighted least-squares estimators based on a single trajectory of population size counts. We prove that, conditional on non-extinction up to a finite time $n$, our estimators are consistent and asymptotic normal as $n\to\infty$. We pay particular attention to estimating the carrying capacity of a population. Our estimators are the first conditionally consistent estimators for PSDBPs, and more generally, for Markov models for populations with a carrying capacity. Through simulated examples, we demonstrate that our estimators outperform other least squares estimators for PSDBPs in a variety of settings. Finally, we apply our methods to estimate the carrying capacity of the endangered Chatham Island black robin population.

We consider population-size-dependent branching processes (PSDBPs) which eventually become extinct with probability one. For these processes, we derive maximum likelihood estimators for the mean number of offspring born to individuals when the current population size is $z\geq 1$. As is standard in branching process theory, an asymptotic analysis of the estimators requires us to condition on non-extinction up to a finite generation $n$ and let $n\to\infty$; however, because the processes become extinct with probability one, we are able to demonstrate that our estimators do not satisfy the classical consistency property ($C$-consistency). This leads us to define the concept of $Q$-consistency, and we prove that our estimators are $Q$-consistent and asymptotically normal. To investigate the circumstances in which a $C$-consistent estimator is preferable to a $Q$-consistent estimator, we then provide two $C$-consistent estimators for subcritical Galton-Watson branching processes. Our results rely on a combination of linear operator theory, coupling arguments, and martingale methods.

We consider the extinction events of Galton-Watson processes with countably infinitely many types. In particular, we construct truncated and augmented Galton-Watson processes with finite but increasing sets of types. A pathwise approach is then used to show that, under some sufficient conditions, the corresponding sequence of extinction probability vectors converges to the global extinction probability vector of the Galton-Watson processes with countably infinitely many types. This gives rise to a number of iterative methods for the computation of the global extinction probability vector., Comment: 35 pages, 11 figures; References 4 and 6 updated

Branching processes are stochastic processes describing the evolution of populations of individuals which reproduce and die independently of each other according to specific probability laws. We consider a particular class of branching processes, called Markovian binary trees, where the lifetime and birth epochs of individuals are controlled by a Markovian arrival process. Our objective is to develop numerical methods to answer several questions about Markovian binary trees. The issue of the extinction probability is the main question addressed in the thesis. We first assume independence between individuals. In this case, the extinction probability is the minimal nonnegative solution of a matrix fixed point equation which can generally not be solved analytically. In order to solve this equation, we develop a linear algorithm based on functional iterations, and a quadratic algorithm, based on Newton's method, and we give their probabilistic interpretation in terms of the tree. Next, we look at some transient features for a Markovian binary tree: the distribution of the population size at any given time, of the time until extinction and of the total progeny. These distributions are obtained using the Kolmogorov and the renewal approaches. We illustrate the results mentioned above through an example where the Markovian binary tree serves as a model for female families in different countries, for which we use real data provided by the World Health Organization website. Finally, we analyze the case where Markovian binary trees evolve under the external influence of a random environment or a catastrophe process. In this case, individuals do not behave independently of each other anymore, and the extinction probability may no longer be expressed as the solution of a fixed point equation, which makes the analysis more complicated. We approach the extinction probability, through the study of the population size distribution, by purely numerical methods of resolution of partial differential equations, and also by probabilistic methods imposing constraints on the external process or on the maximal population size. / Les processus de branchements sont des processus stochastiques décrivant l'évolution de populations d'individus qui se reproduisent et meurent indépendamment les uns des autres, suivant des lois de probabilités spécifiques. Nous considérons une classe particulière de processus de branchement, appelés arbres binaires Markoviens, dans lesquels la vie d'un individu et ses instants de reproduction sont contrôlés par un MAP. Notre objectif est de développer des méthodes numériques pour répondre à plusieurs questions à propos des arbres binaires Markoviens. La question de la probabilité d'extinction d'un arbre binaire Markovien est la principale abordée dans la thèse. Nous faisons tout d'abord l'hypothèse d'indépendance entre individus. Dans ce cas, la probabilité d'extinction s'exprime comme la solution minimale non négative d'une équation de point fixe matricielle, qui ne peut être résolue analytiquement. Afin de résoudre cette équation, nous développons un algorithme linéaire, basé sur l'itération fonctionnelle, ainsi que des algorithmes quadratiques, basés sur la méthode de Newton, et nous donnons leur interprétation probabiliste en termes de l'arbre que l'on étudie. Nous nous intéressons ensuite à certaines caractéristiques transitoires d'un arbre binaire Markovien: la distribution de la taille de la population à un instant donné, celle du temps jusqu'à l'extinction du processus et celle de la descendance totale. Ces distributions sont obtenues en utilisant l'approche de Kolmogorov ainsi que l'approche de renouvellement. Nous illustrons les résultats mentionnés plus haut au travers d'un exemple où l'arbre binaire Markovien sert de modèle pour des populations féminines dans différents pays, et pour lesquelles nous utilisons des données réelles fournies par la World Health Organization. Enfin, nous analysons le cas où les arbres binaires Markoviens évoluent sous une influence extérieure aléatoire, comme un environnement Markovien aléatoire ou un processus de catastrophes. Dans ce cas, les individus ne se comportent plus indépendamment les uns des autres, et la probabilité d'extinction ne peut plus s'exprimer comme la solution d'une équation de point fixe, ce qui rend l'analyse plus compliquée. Nous approchons la probabilité d'extinction au travers de l'étude de la distribution de la taille de la population, à la fois par des méthodes purement numériques de résolution d'équations aux dérivées partielles, ainsi que par des méthodes probabilistes en imposant des contraintes sur le processus extérieur ou sur la taille maximale de la population., Doctorat en Sciences, info:eu-repo/semantics/published

We consider the extinction events of Galton-Watson processes with countably infinitely many types. In particular, we construct truncated and augmented Galton-Watson processes with finite but increasing sets of types. A pathwise approach is then used to show that, under some sufficient conditions, the corresponding sequence of extinction probability vectors converges to the global extinction probability vector of the Galton-Watson process with countably infinitely many types. Besides giving rise to a family of new iterative methods for computing the global extinction probability vector, our approach paves the way to new global extinction criteria for branching processes with countably infinitely many types. (C) 2018 Elsevier B.V. All rights reserved.