User:Carl Boettiger/Notebook/Comparative Phylogenetics/2010/08/27
|Comparative Phylogenetics||Main project page|
Previous entry Next entry
Likelihood ratio ToDo
Power test results
An estimate of how informative a phylogeny is for applications in comparative methods. The independent of trait data, depending only on the structure and size of the tree. (i.e. more taxa help, but unresolved polytomes help less -- a large star tree has zero power). All trees are normalized to unit length for comparison of the weakest strength of selection detectable on the tree. For instance, in the 5-taxa primate tree, even if the strength of selection were strong enough to decrease the correlation between ancestors (by e^-2) along a branch only 1/50th of the whole tree length, one would still have only about a 20% chance of detecting that force (at the 95% confidence level). Meanwhile on the carnivora tree one would have about 70% chance of detecting a force of selection 50 times weaker than that, a force that would take the entire time in the tree to decrease the correlation between ancestors by e^-2.