Multitype $Λ$-coalescents and continuous state branching processes

Adrián González Casanova,Noemi Kurt,Imanol Nuñez Morales,José Luis Pérez
2024-12-31
Abstract:We provide new connections between multitype $\Lambda$-coalescents and multitype continuous state branching processes via duality and a homeomorphism on their parameter space. The approach is based on a sequential sampling procedure for the frequency process of independent CSBPs, and provides forward and backward processes for multitype population models under $\Lambda$-type reproduction. It provides some insight on different approaches to generalise $\Lambda$-coalescents to the multitype setup.
Probability
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to establish the connection between multitype Λ - coalescents and multitype continuous state branching processes (multitype CSBPs). Specifically, by introducing a new method, namely based on homeomorphisms and duality on the parameter space, the authors connect multitype Λ - coalescents with multitype CSBPs and explore the applications of these processes in population models. ### Main problems 1. **Definition and properties of multitype Λ - coalescents**: - The authors extend Pitman's definition of single - type Λ - coalescents to the multitype case. - The characteristics of multitype Λ - coalescents are proposed, including how blocks of different types merge and their time evolution. 2. **Frequency processes of multitype CSBPs**: - By introducing a sequential sampling procedure, the frequency processes of two independent multitype CSBPs are studied. - It is proved that this frequency process can be time - changed to become the dual process of multitype Λ - coalescents. 3. **Homeomorphism and duality**: - The homeomorphism relationship between multitype CSBPs and multitype Λ - coalescents is established. - It is proved that this homeomorphism relationship can be used to describe the genealogy of multitype population models. ### Summary of mathematical formulas To better understand these problems, some key mathematical formulas are listed below: - **Rate of multitype Λ - coalescents**: \[ \lambda_{\mathbf{k},\mathbf{j}}=\Lambda(\{0\})1_{\{\mathbf{j} = 2\mathbf{e}_i\}}+\int_{(0,1]^d}x^{\mathbf{j}-2\mathbf{e}_i}(1 - x)^{\mathbf{k}-\mathbf{j}}\Lambda(d\mathbf{x}) \] where $\mathbf{k}\in\mathbb{N}^d_0$ represents the number of blocks, $\mathbf{j}$ represents the number of merged blocks, and $\mathbf{e}_i$ is the $i$ - th standard basis vector. - **Generator of multitype CSBPs**: \[ \mathcal{L}f(\mathbf{x})=\langle\mathbf{F}\mathbf{x},\nabla f(\mathbf{x})\rangle+\sum_{i = 1}^d\sigma_i x_i\frac{\partial^2 f}{\partial x_i^2}(\mathbf{x})+\sum_{i = 1}^d x_i\int_{\mathbb{R}_+^d}\left(f(\mathbf{x}+\mathbf{y})-f(\mathbf{x})-\langle\mathbf{y},\nabla f(\mathbf{x})\rangle\right)\nu_i(d\mathbf{y}) \] where $\mathbf{F}$ is a matrix, $\sigma_i$ is a diffusion term, and $\nu_i$ is a measure that controls large reproduction events. - **Generator of frequency processes**: \[ \mathcal{A}^{(\mathbf{c})}f(\mathbf{u})=\sum_{i = 1}^d\sum_{j\neq i}F_{ij}c_j\left(u_j - u_i\right)\frac{\partial f}{\partial u_i}(\mathbf{u}) \]