THEORY

Introduction

Throughout the book a number of approaches have been exemplified to assess and compare various aspects of evolutionary trees and models.

To check the reliability of branches in a certain tree, one can use (non-parametric) bootstrapping or jackknifing, combining alignment subsampling and consensus trees to get support values on branches (Chapter 5). Furthermore, other methods that generate or sample sets of plausible trees can be used to get support values, like Bayesian MCMC sampling (Chapter 7) or quartet puzzling (Chapter 6).

Various approaches have been devised to determine a best-suited evolutionary model (Chapter 10). Such approaches are often based on the maximum likelihood values obtained for the models in question. Different measures are applied like Akaike Information Criterion (AIC), Bayesian Infomation Criterion (BIC), Akaike Weights, and other model selection techniques (refer to Johnson & Omland, 2004, for review) to correct for the additional parameters in the more complex models. Such techniques are also implemented in programs like modeltest to select the most useful model of evolution (see Posada & Buckley, 2004 and Chapter 10 for details).

In this chapter we will briefly review different techniques and tests to compare contradicting and, hence, non-nested tree topologies using their likelihood values. Since a large variation of testing approaches can be applied (see, e.g. Goldman et al., 2000), we will restrict ourselves to review a number of common tests for which easily accessible software implementations exist.