[Forum SIS] ammissione esami finali di dottorato

[Elio Mineo]

Cari colleghi,
vi inoltro questo messaggio di Marcello Chiodi.
Cari saluti,
Elio Mineo

------- Messaggio inoltrato -------
Da: Marcello Chiodi <marcello.chiodi at unipa.it>
A: 'Elio Mineo' <mineoeli at unipa.it>
Oggetto: ammissione esami finali di dottorato
Data: Thu, 01 Dec 2011 09:22:38 +0100


Dottorato di ricerca in Statistica, Statistica applicata e Finanza
quantitativa 
del Dipartimento di Scienze Statistiche e Matematiche "Silvio Vianelli" 
Università degli studi di Palermo 

------------------------------------------------------------------------------------------- 
A TUTTI  GLI INTERESSATI 

Il giorno 6 dicembre 2011 alle ore 15 presso il Dipartimento di Scienze
Statistiche e Matematiche "S.Vianelli" di Palermo (Aula "Antonino
Mineo") 
gli studenti del terzo anno del dottorato di ricerca presenteranno le
loro relazioni conclusive  per l'ammissione agli esami finali del
dottorato: 

Dott. Antonio Abbruzzo: 
Graphical models for estimating dynamic networks 

Dott. Salvatore Marcantonio:
Causal models for University of Palermo Ordinary Financing Fund
Dott.ssa Clara Romano:
Student Evaluation Teaching:multilevel IRT model

Il Coordinatore del Dottorato di ricerca in Statistica, Statistica 
applicata e Finanza quantitativa 

Marcello Chiodi 

abstract delle tre relazioni:

Antonio Abbruzzo:

Graphical models for estimating dynamic networks 


Abstract.
Estimating the structures of dynamic networks from data is an active
research area which has many potential applications in various domains,
including molecular biology, social science and marketing data analysis.
For example, discovering gene regulatory networks from microarray is one
important direction in system biology.  

Recently, penalized graphical models were proposed to estimate static
networks in high dimensional studies because of their statistical
properties and computational tractability.


We propose penalized graphical models to estimate structured dynamic
networks, for detecting time evolution of dynamic networks, and to
estimate particular topological structures such as scale-free dynamic
networks for small world structures. These models can be applied when
estimating dynamic networks in high dimensional environments.


The problem of estimating dynamic networks becomes even more challenging
when latent variables are involved in a larger system, i.e. some
components of a network can not be observed. State space models have
been proposed in order to study dynamic networks with latent variables.
However, expectation maximization combined with Kalman filters for
estimating dynamic networks with latent variables can be very unstable.
We propose penalized Gaussian graphical models to estimate dynamic
networks with latent structures. 


When multivariate dynamic data are binary and ordinal random variables,
transformation based on probability distribution with fixed marginals
can be used to do inference. We consider Gaussian copula for non
Gaussian graphical models to overcome the assumption of multivariate
Gaussian data.


Finally, we apply the proposed methodologies to "human T-cell" dataset,
a time-course microarray experiment.


--------------------------------------------------------------------------------------------------
Salvatore Marcantonio:
Causal models for University of Palermo Ordinary Financing Fund

abstract:
University Ordinary Financing Fund (FFO) is the budget primary receipt category. For three years now, an increasing part of the fund is allocated to universities based on a set of performance indicators regarding educational offer and scientific research.
The aim of the thesis is monitoring three of them: educational offer quality (A1),  educational processes results (A2), and scientific research quality through Project of Relevant National Interest (B1).
Monitoring means a four-fold analysis:
Descriptive: tables and plots showing the phenomenon as it is;
Retrospective: answering, using counterfactuals, to question like: what if some past events  had gone differently.
Prevision: using time serie analysis in a bayesian framework predicting one or two year ahead indicators value;
Intervention: using causal and counterfacual analyis according to Judea Pearl methodology, predict what would be the effect of an intevention, e.g. a policy, on the indicators value.


--------------------------------------------------------------------------------------------------
Clara Romano:
Student Evaluation Teaching:multilevel IRT model

The main aim of this thesis is to measure the quality of teaching,
through levels of satisfaction of students, on several aspects of
university courses (i.e. items). 
In the first part of this work, we consider aggregated data in order to
give some suggestions to the policy makers on the variables or items
that determine the students' opinion. In particular, we are interested
to determine an explicit quantication of the relative importanceof each
item for the overall satisfaction of teaching.
In the second part we focus on individual data (i.e. students), so that
we introduce the students characteristics as variables within the model.
Moreover, we assess whether students' characteristics can aect the
teaching evaluation. Summarizing, firstly, we highlight any dierences
in terms of satisfaction among items, using simple statistical tools.
Then, we apply Multilevel Item Response Models to take into account the
complex structure of our data.












-- 
*******************************************
MY NEW PHONE NUMBER: +39 091 23895236
change your address contact with my new mail addrees:
marcello.chiodi at unipa.it
the old address chiodi at unipa.it will still only work for a while 
Prof. Marcello Chiodi

Dipartimento di Scienze Statistiche e Matematiche "Silvio Vianelli"
Università degli studi di Palermo
viale delle Scienze (ed. 13)
90128 Palermo
Italy
tel +39 091 23895236

<http://dssm.unipa.it/chiodi/>

Leggi le pubblicazioni on line della SIS (Società Italiana di
Statistica): SIS-magazine, in aggiornamento continuo
http://magazine.sis-statistica.it/ <http://magazine.sis-statistica.it/>
e Statistica&Società, rivista quadrimestrale di diffusione della cultura
statistica http://dssm.unipa.it/statSoc/ <http://dssm.unipa.it/statSoc/>