H

t=l 1.000-1.700 EZ3 1.700-2.400 CZJ 2.400-3.100 tsa 3.100-3.800 3.800-4.500 4.500-5.200 a 5.200-5.900 ra 5.900-6.600 rn 6.600-7.300 im 7.300-8.000 O Equivocal

t=l 1.000-1.700 EZ3 1.700-2.400 CZJ 2.400-3.100 tsa 3.100-3.800 3.800-4.500 4.500-5.200 a 5.200-5.900 ra 5.900-6.600 rn 6.600-7.300 im 7.300-8.000 O Equivocal

(b)

Figure 3 Alternative methods for estimating ancestral character states. a, Linear parsimony. b, Squared-change parsimony. Character state data by taxon reported in the table.

hypothesis given a model of trait evolution (defined by user). Bayesian analysis (BA) selects the ancestral trait value with the highest posterior probability, given the probabilities of priors (external evidence) and assumptions of trait evolution (defined by user). Because they are model-based approaches, ML and BA optimization methods are more commonly used in the analysis of gene sequence data, using explicit models of changes between nucleotide bases (Lio and Goldman, 1998; Sullivan et al., 1999). ML has been used in the analysis of continuous character evolution where the models may vary from very simple (e.g., Brownian motion) to quite complex; there is a large literature regarding methods to test the validity of using particular models (Diaz-Uriarte and Garland, 1996; Oakley, 2003).

Was this article helpful?

0 0

Post a comment