[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
no subject (file transmission)
Avviso di corso di dottorato:
Gaussian graphical models
Prof. Gerard Letac
Universite' Paul Sabatier
Toulouse, France
Universita' di Roma "La Sapienza"
Dipartimento di Matematica Istituto "Guido Castelnuovo"
Aula B, Tuesday and Wednesday, 14:00 - 16:00, Nov. 8 - Dec. 14
Program
November 8th, Tuesday. Simpson paradox. Discrete
graphical models. A refreshment about Gaussian distributions in
$\mathbb{R}^n$ and the estimation of their parameters.
November 9th, Wednesday. Dempster models.
Why estimation by maximum likelihood is impossible if the graph is
not decomposable. The two fundamental cones $P_G$ and $Q_G$ of a
decomposable graph.
November 15th, Tuesday. The geometry of the decomposable graphs: the
maximal cliques, the minimal separators and their multiplicities.
November 16th, Wednesday. Examples of decomposable graphs: $A_n$,
daisies, trees, homogeneous graphs.
November 22th, Tuesday. The Gavril Buneman theorem:
every decomposable graph is a graph of subtrees of a tree.
November 23th, Wednesday. Perfect ordering of the
vertices, perfect ordering of the maximal cliques of a
decomposable graph.
November 29th, Tuesday. The Lauritzen's magical formula for a
decomposable graph. Application to the duality between the two
cones $P_G$ and $Q_G.$
November 30th, Wednesday. A refreshment on the
classical Wishart distribution. The Dawid-Lauritzen and the
Roverato distributions for a decomposable graph.
December 6th, Tuesday. The Wishart distributions of
type I and II for a decomposable graph. The eigenvalue property.
The case of homogeneous graphs.
December 7th, Wednesday. The Wishart distributions of
type I and II for a decomposable graph: the non homogeneous case.
Explicit computation for $A_4.$
December 13th, Tuesday. An elementary look on the real homogeneous
cones and their use in statistics (S. Andersson's theory).
December 14th, Wednesday. Wishart distributions of type
I and II and their use in Bayesian estimation. Their hyperMarkov
properties.
For more information, please contact:
Prof. Mauro Piccioni (e-mail: piccioni@mat.uniroma1.it)
=============================================================================
NOTA:
Le norme per utilizzare il forum SIS e le istruzioni per iscrizione
e cancellazione sono disponibili all'indirizzo
.
. http://w3.uniroma1.it/sis/forum.asp
.
L'archivio di tutti i messaggi (aggiornato al giorno precedente)
e' disponibile all'indirizzo
.
. http://www.stat.unipg.it/cgi-bin/wilma/sis/
.
=============================================================================