[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Seminari di probabilita' e statistica al Politecnico di Torino




SEMINARI DI PROBABILITA' E STATISTICA AL POLITECNICO DI TORINO

Martedi' 4 febbraio 2003
Aula Buzano, Dipartimento di Matematica, Politecnico di Torino
Corso Duca degli Abruzzi, 24
10129 Torino


ore 14:30
ANTONIETTA MIRA
Universita' dell'Insubria, Varese
"An introduction to MCMC, Reversible Jumps and Delaying Rejection
algorithms"

Markov chain Monte Carlo (MCMC) is a method to estimate complicated
integrals of functions with respect to a
possibly multidimensional distributions. A Markov chain, stationary with
respect to the distribution of interest, is
run producing a sample path. The integral is approximated by taking the
average of the function over the sample. In
this talk we present the standard MCMC algorithm, the Metropolis-Hastings
algorithm and possible ways to
generalize it (Reversible jumps) and improve its efficiency (Delaying
Rejection). Examples of applications will be
discussed.


ore 16:00
LUISA SCACCIA
Universita' di Perugia
"Bayesian inference for latent class model with application to the
analysis of capture-recapture data"

A well known method for estimating the size N of a certain population is
the capture-recapture method. A typical
capture-recapture study consists in capturing and somehow marking subjects
at different occasions so that a
capture configuration may be associated to each subject captured at least
once. The problem is then to estimate
the probability of never being captured and, consequently, to estimate N.
A recent approach to the problem is
based on the classical latent class model. This model assumes that the
population may be divided into C classes so
that the subjects within each class have the same degree of catchability
with respect to the same trapping
occasion. We formulate the latent class model in a Bayesian framework
avoiding, in this way, the problem of
arbitrary fixing the number of the latent classes. We use the Reversible
Jump algorithm to sample from the
posterior distribution of N and from such a posterior, we obtain point
estimates and credibility intervals for N.


Tutti gli interessati sono invitati a partecipare.

Il programma generale dei seminari di probabilita' e statistica al
Politecnico di Torino e' disponibile all'indirizzo
http://calvino.polito.it/~probstat/seminars.html

Cordiali saluti,

		Federica Giummole'


************************************************************************
Federica Giummole'
Dipartimento di Matematica     Tel: +39 011 564 7550
Politecnico di Torino          Fax: +39 011 564 7599
Corso Duca degli Abruzzi 24    email: giummole@calvino.polito.it
10129 Torino                   www: http://calvino.polito.it/~giummole/
ITALY
************************************************************************