|
Abstract:
|
Morphological evolutionists make phylogenies for characters that have discrete states, such as 0/1 characters. They usually do so using a nonstatistical "parsimony method" that minimizes the number of changes of state of the character along the tree. There has been some attempt to make statistical methods that ask whether two of these characters change along a tree in a correlated way. Papers by Pagel and Lewis have used a simple two-state Markov process to ask about correlated change and to infer phylogenies. This talk will emphasize a more realistic alternative, the threshold model of quantitative genetics, due to Sewall Wright (1934). An underlying continuous character, the "liability," wanders on a continuous scale, a developmental threshold converts this to the observed 0/1 character. Covariances of the underlying characters can be inferred using a MCMC method given a tree.
|
