Some Poisson-based processes at geometric times (Q6110781)

The authors of \textit{E. Orsingher} and \textit{F. Polito} [J. Stat. Phys. 148, No. 2, 233--249 (2012; Zbl 1251.60041)] considered the composition of a homogeneous Poisson process to an independent Poisson process, both homogenous and inhomogeneous. They focused on establishing some distributional relations between these composed processes and random sums. In [\textit{A. Di Crescenzo} et al., J. Appl. Probab. 52, No. 2, 360--374 (2015; Zbl 1323.60066)], instead, a compound Poisson process whose running time is an independent Poisson process is studied. In [\textit{L. Beghin} and \textit{E. Orsingher}, J. Stat. Phys. 163, No. 1, 1--21 (2016; Zbl 1338.60131)], mainly interested in Poisson random times, the authors investigated, among others, linear birth processes, linear and sublinear death processes with an independent Poisson subordinator, with an emphasis on the hitting probabilities. As a consequence, it becomes necessary to explore potential alternatives to a Poisson process. A fair compromise which preserves mathematical manageability is represented by a geometric counting process [\textit{J. H. Cha} and \textit{M. Finkelstein}, Probab. Eng. Inf. Sci. 27, No. 2, 177--185 (2013; Zbl 1276.60058)]. A geometric counting process with parameter \(\mu >0\) is the process \(G_{\mu} = \{G_{\mu}(t), t \ge 0\}\), whose one-dimensional distribution can be expressed, for \(k\in \{0,1,2,\dots\}\), as \(P(G_{\mu}(t) = k) = \int_0^{+\infty}P(N_{\lambda}(t)= k)\,dU_{\mu}(\lambda)\), \(t \ge 0\), where \(N_{\lambda} = \{N_{\lambda}(t), t \ge 0\}\) is a Poisson process with rate \(\lambda> 0\), and where \(U_{\mu}(\cdot)\) is the exponential distribution with mean \(\mu>0\). The geometric counting process is a particular case of a much wider class of well-known processes, the mixed Poisson processes, which means that the intensity of the Poisson process is regarded as a random variable. For a complex variable \(y\), the geometric polynomials \(\omega_n(y)\) of degree \(n\) are defined by \(\omega_n(y) = \sum_{k=0}^n \{n, k\} k! y^k\), \(n\in \{0,1,2,\dots\}\), where \(\{n, k\} = \frac{1}{k!} \sum_{i=0}^k (-1)^i \frac{k!}{i! (k-i)!} (k-i)^n\), \(k=0,1,2\dots,n\), are the Stirling numbers of the second kind. In the present paper three Poisson-based stochastic processes whose directing time is a geometric process are studied. The aim is to provide new models for the description of both count and continuous data closer to reality than existing ones, but still appealing when it comes to computing. The paper is organized as follows. Section 2 contains definitions and preliminaries on the geometric counting process and on the geometric polynomials. As a novelty, the author proves two propositions regarding the summation of series of geometric polynomials which generalize the generating function. In Section 3 the author investigates a homogeneous Poisson process directed by an independent geometric process. The probability distribution of the derived process, the study of its main properties and its connection with compound Poisson-geometric random variables are obtained. The author also focus on the distribution of the first-passage time through a constant boundary. Sections 4 and 5 are devoted to the analysis of a compound Poisson process and a multiplicative compound Poisson process respectively, both at independent geometric times. In either case, after obtaining the general expression for the probability law of the resulting processes, the author specifies the distribution of the jump size. This is noteworthy, since writing in closed form the law of a random sum or a random product is very rare. Section 6 concludes the paper with a discussion on future research directions. These mainly include, but are not limited to, the application of the proposed models to the study of the behaviour of certain physical and biological systems.