************** AVVISO DI SEMINARIO
***************
Nell'ambito del progetto di ricerca europeo
"Bayesian networks for forensic inference from genetic markers".
Giovedi' 5 giugno, ore 17,00, Sala del Consiglio
(accanto alla Presidenza)
Nuala A. Sheehan,
Department of Epidemiology and Public Health,
University of Leicester
terrą presso il Dipartimento di Economia, Universitą Roma Tre, Via Ostiense
139, un seminario dal titolo
ON THE PERFORMANCE OF A BLOCKING MCMC SAMPLER IN A SIMPLE
GENETIC MAPPING SCENARIO ABSTRACT
It is well known that algorithms for performing exact probability
and likelihood computations on pedigrees break down in practice for complex genetic applications. Markov chain Monte Carlo (McMC) methods can be employed to provide estimates of the required quantities but although these methods are not restricted by the complexity of the problem, the estimates they yield are often unreliable. With this in mind, we investigate the behaviour of one particular McMC sampling scheme for the detection of a quantitative trait locus by linkage with marker data when a few specific aspects of the problem change. Although we have chosen a very simple pedigree structure and a trivial mapping scenario, the problem is sufficiently complicated that exact methods no longer apply. Consequently, in order to assess the efficiency of our sampler, we have carried out a study on a test-bed of simulated datasets. One question that should always be of interest in these complex applications is whether a full Bayesian McMC analysis is sufficiently better than a simpler approximation as to make it worth the extra computational effort required. ____________________________________________________________________
Tutti gli interessati sono invitati a
partecipare. _____________________
Paola Vicard Dipartimento di Economia Universitą Roma Tre Via Ostiense 139 00154 Roma Tel.: +39 06 57374219 Fax: +39 06 57374093 e-mail: vicard@uniroma3.it _____________________ |