Success of maximum likelihood phylogeny inference in the four-taxon case.

摘要: We used simulated data to investigate a number of properties maximum-likelihood (ML) phylogenetic tree estimation for the case four taxa. Simulated were generated under broad range conditions, including wide variation in branch lengths, differences ratio transition and transversion substitutions, absence presence gamma-distributed site-to-site rate variation. Data analyzed ML framework with two different substitution models, we compared ability models reconstruct correct topology. Although both inconsistent some branch-length combinations variation, efficient predictors topology most simulation conditions. also examined performance likelihood (LR) test significant positive interior length. This was found be misleading many rejecting too often Under null hypothesis zero length internal branch, LR statistics are assumed asymptotically distributed chi 2(1); limited data, distribution varies from 2(1).