The Bell-Touchard Counting process

TL;DR

This paper introduces the Bell-Touchard process, a new counting process based on the Bell-Touchard distribution, extending Poisson processes to model count data without the rare events restriction, with potential applications in complex stochastic modeling.

Contribution

The paper proposes the Bell-Touchard process, a novel counting process inspired by discrete distributions, and explores its properties, generalizations, and natural emergence from Poisson process compositions.

Findings

The Bell-Touchard process is a compound and multiple Poisson process.

It is closed under convolution and decomposition operations.

The process arises from the composition of two Poisson processes.

Abstract

The Poisson process is one of the simplest stochastic processes defined in continuous time, having interesting mathematical properties, leading, in many situations, to applications mathematically treatable. One of the limitations of the Poisson process is the rare events hypothesis; which is the hypothesis of unitary jumps within an infinitesimal window of time. Although that restriction may be avoided by the compound Poisson process, in most situations, we don't have a closed expression for the probability distribution of the increments of such processes, leaving us options such as working with probability generating functions, numerical analysis and simulations. It is with this motivation in mind, inspired by the recent developments of discrete distributions, that we propose a new counting process based on the Bell-Touchard probability distribution, naming it the Bell-Touchard process. We verify that the process is a compound Poisson process, a multiple Poisson process and that it is closed for convolution plus decomposition operations. Besides, we show that the Bell-Touchard process arises naturally from the composition of two Poisson processes. Moreover, we propose two generalizations; namely, the compound Bell-Touchard process and the non-homogeneous Bell-Touchard process, showing that the last one arises from the composition of a non-homogeneous Poisson process along with a homogeneous Poisson process. We emphasize that since previous works have been shown that the Bell-Touchard probability distribution can be used quite effectively for modelling count data, the Bell-Touchard process and its generalizations may contribute to the formulation of mathematical treatable models where the rare events hypothesis is not suitable.