STATISTICAL TESTS OF PHENOGRAMS BASED ON GENETIC DISTANCES.

Methods used for estimating the confidence levels of UPGMA-type phenograms based on molecular-genetic data are examined. The methods based on internodal variances or on bootstrapping over characters are compared by simulating a three-species phylogeny with a trifurcation. The bootstrap method seems slightly better in this comparison. Weighting of OTU's when constructing the phenogram is also examined. A method that weights each OTU according to the estimated independent evolutionary information, a modified WPGMA, appears slightly better than UPGMA in estimating the branching points and branch lengths. The methods are applied to the data on restriction sites in eight Hawaiian Drosophila species. The significance of the clusters among them differ from those reported earlier.