Adequacy of numerical taxonomic methods

Abstract

A simulation model, GENESIS, was developed to examine the relative merits of several phylocenetic and phenetic methods. For a description of the model and the results of a sensitivity analysis, see Heijerman (1988, 1990). GENESIS was designed to generate artificial data sets of ‘species’ with known phylogenies. These datasets were subjected to character analysis b various numerical taxonomic methods (UPGMA clustering, Wagner parsimony analysis and compo nent-comparibility analysis). The results of analysis were compared with the true phylogeny. The agreement between the true tree and the reconstructed tree was used as a measure of quality (adequacy). By varying the input parameters of GENESIS, output produced under different evolutionary scenarios was obtained and the relative adequacies of the methods in relation to these evolutionary conditions were evaluated. The overall differences in adequacy between Wagner parsimony as performed by PAUP, PHYLIP (MIX) and HENNIG86, and UPGMA clustering with product-moment correlations of standardized characters, were rather small. These methods were more adequate than Wagner parsimony with WAGNER78 and group com patibility with CAFCA. The adequacy of the trees as estimated by WAGNER78, PAUP, PHYLIP and CAFCA depended on several tree properties, the consistency index being the most important one.