AN EMPIRICAL COMPARISON OF MICROCOMPUTER PARSIMONY PROGRAMS.

Abstract

Abstract- The effectiveness and efficiency of two microcomputer parsimony programs (Felsenstein's PHYLIP and Swofford's PAUP), and a prototype version of a third (SHEN, to be incorporated in Farris' HENNIG-86), are evaluated with reference to 35 data sets, including those previously used by Luckow and Pimentel (1985) to benchmark mainframe programs. Both PHYLIP and PAUP can be used effectively; with careful selection of options, their accuracy can equal or surpass that of older mainframe programs. PHYLIP is relatively inefficient (in use of computer time); its usefulness is also limited by the inability of its heuristic algorithms to detect multiple equally parsimonious solutions in a single run. PHYLIP's exact algorithm (branch-and-bound) and PAUP's heuristic and exact algorithms do not share the latter drawback but require users to spend unreasonable amounts of time coping with redundant output. The remarkable effectiveness of these programs offers hope that (particularly with the advent of multitasking microcomputers) exact and cost-effective solutions will be obtainable in many, if not all, cladistic analyses.