Divergence rates for explosive birth processes (Q1198555): Difference between revisions

This paper considers a pure birth process, \(N_ t\), in which the waiting time in \(j\) is \(X_ j\). The waiting times are independent random variables. The explosive case, in which the \(X_ j\) have a finite sum, is of interest. It is noted that if \(X_ j/EX_ j\) are independent identically distributed variables, then the condition that \(\sum EX_ j<\infty\) is necessary and sufficient for the process to explode, but that this condition is not necessary in general. Let the time of the first infinity be \(T_ \infty\) and let \(\psi(j)\) be the tail sum, from \(j\) onwards, of \(\{EX_ i\}\). A special case of the main result is that, under suitable conditions, \(\psi(N_ t)\sim T_ \infty-t\) as \(t\uparrow T_ \infty\). Using this, a variety of examples are constructed. The motivation was an example in a paper by \textit{D. A. Dawson} and \textit{K. Fleischmann} [J. Appl. Probab. 26, No. 1, 9-22 (1989; Zbl 0679.60074)] in which \(N_ t\) looks like \((T_ \infty-t)^{-1}\).