A0546
Title: A new approach for estimating the largest mean in a Gaussian mixture model with applications in population genetics
Authors: Andreas Futschik - JKU Linz (Austria) [presenting]
Abstract: A new method is proposed to estimate the mixture component with the largest mean parameter when the data come from a Gaussian mixture model. Some properties of the method are discussed, and it is shown that it has advantages compared to classical approaches of inference, such as the EM algorithm, when there are many components. The method relies on inference for the truncated normal distribution. Our motivating application comes from population genetics, where the effective population size $N_e$ is an important parameter when specifying null models. It first is explained how $N_e$ has usually been estimated. Then it is shown how our proposed method may be used to identify the neutral $N_e$.
