Efficiencies of different statistical tests in supporting a known vertebrate phylogeny.

In a recent paper (Russo, Takezaki, and Nei 1996) we examined the efficiencies of different tree-building methods (neighbor joining [NJ], minimum evolution, maximum parsimony [MP], and maximum likelihood) in recovering a true vertebrate phylogeny using mitochondrial genes. In practice, however, we usually do not know the true tree, so the reliability of a reconstructed tree is tested by various statistical methods. One of the most commonly used statistical methods for testing the reliability of phylogenetic trees is the bootstrap test of each interior branch (Felsenstein 1985). Even though this test seems to be conservative in theorectical studies (e.g., Sitnikova, Rzhetsky, and Nei 1995), there is little evidence of how well it performs empirically. value used in the standard statistical test. This computation was done by the METREE program (Rzhetsky and Nei 1994) that calculates the CP values for NJ trees.