************** AVVISO DI SEMINARIO
***************
Martedì 26 marzo, ore 16,00 Aula 10
La Dott.ssa Vanessa Didelez, Department of
Statistical Science, University College London
terrà presso il Dipartimento di Economia, Università
Roma Tre, Via Ostiense 139, un seminario dal titolo
GRAPHICAL MODELS FOR EVENT HISTORY DATA
ABSTRACT
This talk deals with the application
of graphical models to
dynamic data, in particular to event history data. The
"classical"
graphical models represent conditional independencies among
components of
a multivariate random vector by a suitable graph. Several
approaches have
been proposed to generalise this basic idea for the case that
a time
component has to be taken into account, the challenge being to
adequately
represent dynamic dependencies. For event history analysis this
has led me
to drop the concept of conditional independence in favour of the
so-called
'local independence'. Heuristically, an event A is locally
independent of
another event B if the intensity for the former is not altered
by the
occurrence of the latter. One of the most important differences
to
conditional independence is the asymmetry of local
independence,
reflecting the dynamic nature of dependencies among events in
time.
I will give a brief introduction to graphical models as well as
an
overview over the existing generalisations for dynamic data. The main
part
of the talk will discuss the concept of local independence and
its
graphical representation. Among others, it is shown that analogous
results
about separation and collapsibility, as known for conditional
independence
graphs, can be obtained for local independence
graphs.
____________________________________________________________________
Tutti gli interessati sono invitati a
partecipare.