The supercritical birth, death and catastrophe process: Limit theorems on the set of non-extinction (Q2641012)

The linear birth and death process with super-imposed linear catastrophe/emigration component is defined as Feller process \((X_ t\), \(t\geq 0)\) on \(N_+\) with generator \(q_{ij}=i\rho b_{i-j}\) \((1\leq j\leq i+1\), \(i\neq j)\), \(q_{i0}=i\rho \sum_{k\geq i}b_ k\), \(q_{ii}=- i\rho\), where \(\rho >0\), \(\{b_ j\), \(j\geq -1\}\) is a discrete distribution satisfying \(b_{-1}>0\), \(b_ 0=0\), \(b_ j>0\) for some \(j\geq 1\). Catastrophes are allowed to occur at a rate proportional to the instantaneous population size by removing j individuals with a probability depending only on j. The present paper improves some results of the author [ibid. 25, 307-325 (1987; Zbl 0642.92012)] in the supercritical case where \(D=-\sum_{j\geq -1}jb_ j>0\). Without any extra moment condition almost sure convergence of \(n_ tX_ t\) to a random variable W is proved, where \(n_ t\) are functions such that \(n_ t\to 0\) as \(t\to \infty\). The distribution of W is a finite mixture of gamma distributions with one atom, at the origin. The norming constants can be expressed as \(n_ t=\exp (-\lambda t)M(\exp \lambda t)\), where M is a slowly varying function. This identification requires estimates of the extinction probabilities as functions of the initial population size. A further result is the weak convergence \([X_ t-W/n_ t]X_ t^{- 1/2}\Rightarrow N(0,\sigma^ 2)\) conditionally on non-extinction \((W>0)\) as \(t\to \infty\) if the moment assumption \(\sum_{j\geq 1}j^ 4b_ j<\infty\) holds.