PHYLOGENETIC ANALYSES OF THE CORRELATED EVOLUTION OF CONTINUOUS CHARACTERS: A SIMULATION STUDY

摘要: We use computer simulation to compare the statistical properties of several methods that have been proposed for estimating evolutionary correlation between two continuous traits, and define alternative correlations may be interest. focus on Felsenstein's (1985) method some variations it "minimum evolution" (of which procedure Huey Bennett (I9871 is a special case), as compared with nonphylogenetic correlation. The last, simple trait values across tips phylogeny, virtually always yields inflated Type I error rates, relatively low power, poor estimates correlations. therefore cannot recommend its use. In contrast, (1 985) acceptable significance tests, high good what we term input standardized realized correlation, given complete phylo- genetic information knowledge rate mode character change (e.g., gradual proportional time ("Brownian motion") or punctuational, only at speciation events). Inaccurate branch length affect any adversely, but rarely does cause perform worse than do others tested. Other generally yield rates lower power. However, certain min- imum evolution (although not specific used by (1987)) often provide more accurate unstandardized their recommended when estimation this desired. also demonstrate how correct can obtained reference an empirical null distribution derived from simulations, practical suggestions choosing analytical method, based both interest availability lengths model appropriate characters being analyzed. Computer programs implement various will simulate (correlated) along known phylogeny are available authors request. These test effectiveness new might proposed, check generality our conclusions regard other phylogenies.