Does the T-PTP test tell us anything we want to know?

Archie (1989), Faith (1990), and Faith and Cranston (1991) devised a test for phylogenetic structure in character data sets, which Faith and Cranston called the permutation tail probability (PTP) test. Faith (1991) introduced a reŽnement intended to test the strength of support for monophyly of individual clades, which he called the topologydependent permutation tail probability (T-PTP) test. Faith presented a priori and a posteriori forms of the test; here, I will discuss the a priori test. All phylogenetic permutation tests involve comparing somepropertyof the originaldata set with a null distribution constructed by randomly scrambling character states within each character across taxa, generating a large number of replicate data sets. The randomized data sets have some of the properties of the real data set—particularly the number of taxa with a given state for each character— but they certainly have no phylogenetic information. The test property used by the a priori T-PTP test for monophyly is the Bremer (1994) support or decay index: the length of the most-parsimonious tree lacking the clade of interest minus the length of themost-parsimonious tree having the clade. The Bremer index of a clade for the original data set is compared with the distribution of Bremer indices of that clade over many randomized data sets. The P-value is simply the proportion of randomized data sets for which the clade’s Bremer index is as great as or greater than its Bremer index for the original data set.