University of Idaho
Department of Mathematics Colloquia

Thursday, February 21, 2008

3:30 p.m. in TLC 031
Refreshments in Brink 330 at 3:00

Efficient computations of Markov functionals lead to novel statistical procedures in evolutionary biology

by

Vladimir Minin
Department of Statistics
University of Washington

Abstract:

In genetics and evolutionary biology, mutations are often modeled as continuous-time Markov chains. In practice, these Markov chains are observed only partially through a finite sample of genetic sequences. However, statistical inference and hypothesis testing often require calculating probabilistic properties of unobserved functionals of Markov evolutionary trajectories. I will discuss efficient numerical algorithms for two of such functionals, evolutionary counting and reward processes. I will demonstrate how these algorithms can be applied to robust estimation of genetic distances, testing co-evolution of two genetic traits, and mapping mutations onto evolutionary histories (phylogenies). I will also describe how Markov functionals can be used to detect and remedy model misspecification in evolutionary applications.