Asymptotic rates of growth of the extinction probability of a mutant gene (Q1193521)

The purpose of the paper is to examine the validity of the relation \[ 1- q\sim 2\varepsilon/\sigma^ 2\quad\text{as}\quad\varepsilon\to 0\tag{\(*\)} \] in a greater generality, where \(q\) is the extinction probability of a branching process with Poisson-distributed offspring distribution with expectation \(1+\varepsilon\) and variance \(\sigma^ 2\). Thereby \(\varepsilon\) denotes the selection advantage of a mutant gene. The paper has been motivated by a paper of \textit{I. Eshel} [ibid. 12, 355- 362 (1981; Zbl 0464.92013)], who gave two counterexamples to equation \((*)\) and proved \((*)\) under certain modified branching conditions. Therefore in section 2 of the paper Eshel's conditions are presented with a discussion of the directions of possible generalizations. Given a family \(\{Z_ n(\varepsilon)\}\) of supercritical Bienaymé-Galton- Watson branching processes a simple necessary and sufficient condition for \((*)\) is extracted in the following, establishing the validity of \((*)\) in those cases where the variance is continuous and strictly positive at \(\varepsilon=0\). The proof is based on the analytic machinery on the remainder term of a Taylor expansion of the offspring distribution. Five examples are given where the expectation and variances are not continuous at zero and yet \((*)\) holds, indicating that \((*)\) is more widely true. Furthermore, a sufficient condition for \((*)\) based on the second factorial moment of the offspring distribution is given. In addition, the nonlinear behavior is discussed indicating that only slight changes in the sequence of probability generating functions result in different rates of convergence for \(1-q\).