PIMS/CSC Distinguished Speaker Series: David Bryant

Abstract:

Simulation experiments are used throughout the sciences to make model comparisons, predict future events, and lobby for particular hypotheses or metholodogies. Simulations can transform subjective design decisions into seemingly objective results. One of the key design decisions is the choice and range of values used for model parameters. We decribe a Markov chain Monte-Carlo (MCMC) framework for conducting simulation experiments. It uses MCMC to sample parameter values in the simulations, thereby improving efficiency and (potentially) transparency. We illustrate the framework with applications to phylogenetics, epidemiology, and archaeology. This is joint work with Jessica Leigh, also at Otago.