Although many methods for inferring species trees from gene trees while taking into account gene tree/species tree discordance provide consistent estimates of the species tree topology, many do not estimate branch lengths or are computationally slow. Exceptions include the GLASS method of Mossel and Roch (2010), the STEAC method of Liu et al. (2009), and the Shallowest Divergence method of Maddison and Knowles (2006). Given estimated gene trees for a set of loci, these methods estimate the divergence time between a pair of taxa by computing either the mean or minimum interspecific coalescence time between the taxa at each locus, and then taking either the mean or minimum of these times across all loci. The species tree is then constructed by applying a suitable hierarchical clustering method to the estimated divergence times. These methods estimate both species tree topologies and branch lengths and are computationally fast. However all of these methods systematically overest imate divergence times. We derive corrections that can substantially reduce the bias and mean squared error in the estimates of divergence times made by these methods and by a fourth method based on the same principle.
view more