Le vendredi 1er octobre 2010, 15 h 30, McGill, Burnside Hall, 805 Sherbrooke O., salle 1B24

A Stopping Rule for MCMC Clustering
Vahid Partovi Nia (McGill University)

RESUME :
When a Bayesian model is formulated for clustering, often Markov chain Monte Carlo (MCMC) method is applied to sample allocations. Therefore a measure of convergence defined on the allocation space, a finite state space, is needed. Such a stopping rule can also be used to quantify efficiency of a chain. A Pearson-like goodness of fit statistic is introduced for Bayesian models with analytically tractable marginal posteriors. The asymptotic distribution of the statistic is derived under equilibrium providing a statistical significance test. Application of the proposed method is demonstrated on MCMC clustering of high-dimensional-low-sample-size metabolite data.