Building on work presented previously (Parker et al., 2008), we study a number of more complex measures of phylogeny-trait association (implemented in the program Befi-BaTS / BaTS v0.10.1) which take into account the branch lengths of a phylogenetic tree in addition to the topographical relationship between taxa. Extensive simulation is performed to measure the Type II error rate (statistical power) of these statistics including those introduced in Parker et al. (2008), as well as the relationship between power and tree shape. The technique is applied to an empirical hepatitis C virus data set presented by Sobesky et al. (2007); their original conclusion that compartmentalization exists between viruses sampled from tumorous and non-tumorous cirrhotic nodules and the plasma is upheld. The association index (AI), migration (PS), phylodynamic diversity (PD) and unique fraction (UF) statistics offer the best combination of Type I error and statistical power to investigate phylogeny-trait association in RNA virus data, while the maximum monophyletic clade size (MC) and nearest taxon (NT) statistics suffer from reduced power in some regions of tree space.
Keywords: BaTS, hepatitis C virus, Markov-chain Monte Carlo, Phylogeny-trait association, Phylogenetic uncertainty, simulation.
Manuscripts in progress (all rights reserved – you may not copy or distribute these files; content and conclusions subject to change; strictly embargoed until publication in a peer-reviewed journal/book):