The survival probability of a mutant in a multidimensional population (Q1193522)

A quite general result is proven concerning the asymptotic behaviour of the survival probability of a slightly supercritical family of \(d\)-type Galton-Watson processes: \[ 1-q(\varepsilon)\sim 2(\rho(\varepsilon)-1) K^{-1} u\quad\text{as }\varepsilon\to 0, \] \(q\) being the extinction probability, \(\rho\) the maximal eigenvalue, \(u\) the right normalized eigenvector of the expectation matrix and \(K\) a constant depending on the variance. Furthermore, the processes depend on a small parameter \(\varepsilon>0\), that we can interpret e.g. in genetic applications as the selection advantage of a mutant gene. The result is a complete analogue of a result of \textit{W. J. Ewens} [J. R. Stat. Soc., Ser. B 30, 164-175 (1968)] for Poisson branching processes. The five step proof of the theorem is based on two series expansions for multidimensional probability generating functions due to \textit{A. Joffe} and \textit{F. Spitzer} [see J. Math. Anal. Appl. 19, 409-430 (1967; Zbl 0178.195)].