Ciclo de SeminÃrios - UnB

[antonio eduardo gomes <aegomes@unb.br>]

           CICLO DE SEMINÃRIOS
PROGRAMA DE PÃS-GRADUAÃÃO EM ESTATÃSTICA
        UNIVERSIDADE DE BRASÃLIA

Data: 30/11/09
HorÃrio: 8:30
Local: AuditÃrio do Departamento de MatemÃtica

Palestrante: Prof. Jose Miguel Bernardo
             Universitat de Valencia, Espanha

TÃtulo: Natural Induction: A Bayesian Approach

Resumo:
A classical statistical problem in the sciences, often referred to
as  the problem of natural induction, is that of determining the
probability that all elements in a finite population have a particular
property, given that all n previous elements from a random sample from
that population possessed the property.
An specific instance of this problem occurs when a zoologist has
observed a number of animals in a secluded habitat, all of which
presented a particular genetic modification, possibly related to
their adaptation to their habitat, and she requires the probability
that all animals in the area have that modification.
The paper on which this talk is based uses new developments in
reference prior theory to propose an explicit, attractive solution
to this problem from an objective Bayesian viewpoint.





----- Final da mensagem encaminhada -----