Is the dying individual the oldest ? (Q1822135)

Consider a population in which the ith individual is born at time \(B_ i\), lives for \(S_ i\) and then dies, at time \(D_ i=B_ i+S_ i\). It is stipulated that the \(B_ i\), \(-\infty <i<\infty\), constitute a Poisson process with rate \(\lambda\), that the \(S_ i\) are i.i.d. with distribution G and that \((B_ i)\) and \((S_ i)\) are independent. Thus, both the point process with points \((B_ i,S_ i)\) and that with points \((D_ i,S_ i)\) are Poisson processes on \({\mathbb{R}}\times {\mathbb{R}}_+\) with mean \(\lambda\ell \times G\), where \(\ell\) denotes Lebesgue measure. The authors study the point process on \({\mathbb{R}}\) whose points \(D^*_ i\) are those death times at which the dying individual is the oldest among those currently surviving (i.e., is the senior member of the population). Using Palm distributions, the (stationary) probability p that the dying individual is the senior is calculated: \[ p=\int^{\infty}_{0}\exp [-\rho (1-G_ e(t))]dG(t), \] where \(\rho =\lambda E[S]\) and \(G_ e(t)=\int^{t}_{0}[1- F(u)]du/\int^{\infty}_{0}[1-F(u)]du\). The process Y whose value at t is the age of the current senior is shown to satisfy \(P\{Y(0)\leq t\}=\exp [-\rho (1-G_ e(t))]\). Other, similar computational results are presented.