Title
Promises and Pitfalls of MCMC for Evolutionary Trees
Speaker
Mary Kuhner, University of Washington
Abstract
Markov chain Monte Carlo (MCMC) methods which search the space of evolutionary trees have become suddenly important in evolution and population biology. They are used both to infer phylogenies (MrBayes) and to estimate population-genetics parameters (LAMARC, BEAST, IM, etc.). In this talk I will describe evolutionary-tree MCMC approaches and talk about their strengths and weaknesses, including several unsolved problems. How can we know whether the search is succeeding? How can we keep it from becoming stuck, especially in cases where we are varying strongly correlated aspects of our model?
Meet the speaker in Room 212 Cockins Hall at 4:30 p.m. Refreshments will be served.
