Convergence of one-dimensional diffusion processes to a jump process related to population genetics (Q809926)

\textit{J. H. Gillespie} [Am. Nat. 121, 697-708 (1983), and Theor. Popul. Biol. 23, 202-215 (1983; Zbl 0507.92010)] conjectured that a K-1 dimensional diffusion process with infinitesimal generator \[ 2^{- 1}\sum^{K-1}_{i,j=1}a_{ij}(x,\alpha_{\epsilon})\partial^ 2/\partial x_ i\partial x_ j+\sum^{K-1}_{i=1}b_ i(x,\alpha_{\epsilon},\theta_{\epsilon})\partial /\partial x_ i \] (\(\epsilon\),\(\alpha {}_{\epsilon},\theta_{\epsilon}>0)\) converges weakly to a Markov chain as \(\epsilon \to 0\) under the condition that \(\alpha_{\epsilon}\to \infty\), \(\theta_{\epsilon}\to 0\). The limiting Markov chains are characterized in terms of \(\lim_{\epsilon \to 0}a_{ij}(x,\alpha_{\epsilon})/\epsilon\) and \(\lim_{\epsilon \to 0}b_ i(x,\alpha_{\epsilon},\theta_{\epsilon})/\epsilon\). Although weak convergence does not hold, the finite-dimensional distributions do converge as it is shown in this paper using limit theorems on bi- generalized diffusion processes. Strong selection - weak mutation limit, moderate selection - weak mutation limit, and weak selection - weak mutation limit are considered for several typical diffusion models of population genetics.