BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Reversible jump MCMC - Eleni Bakra\, MRC Biostatistics Unit
DTSTART:20100203T163000Z
DTEND:20100203T180000Z
UID:TALK23100@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:Reversible jump Markov chain Monte Carlo (RJMCMC) is an extens
 ion of the\nMetropolis-Hastings algorithm for stationary distributions of 
 variable\ndimension. It has been successfully applied in a wide variety of
  settings\,\nincluding the challenging problem of determining the number o
 f components in\na finite mixture and determining the number of states in 
 a hidden Markov\nmodel. In this talk\, RJMCMC is considered as an approach
  to Bayesian model\nselection problems and for this reason\, it is used to
  explore the sampling\nspace that consists of several models of different 
 dimension. The RJMCMC\nalgorithm usually considers a selection of move typ
 es\, some of which explore\nthe parameter space within a model\, and other
 s which propose changes to the\ndimensionality of the model. The choice of
  the proposal mechanism is crucial\nto the performance of the algorithm. T
 herefore\, several methods have been\nproposed in the literature on how to
  choose the proposal mechanism of the\nalgorithm. In this talk\, I will de
 scribe some of these methods and I will\nalso introduce a new one.\n\nThe 
 original reversible jump MCMC paper can be found "here":http://www.jstor.o
 rg/stable/pdfplus/2985194.pdf
LOCATION:MR5\, CMS
END:VEVENT
END:VCALENDAR