Download The activities during my visit at BIGSSS entailed

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

History of the social sciences wikipedia , lookup

Land-use forecasting wikipedia , lookup

Anthropology of development wikipedia , lookup

Neural modeling fields wikipedia , lookup

Economic model wikipedia , lookup

Transcript
Activity report of visit to InGRID
research infrastructures
Name and last name
Federica Nicolussi
Project title
Multivariate Measurement of Poverty Through Chain Graphical Models
(MMPTCGM)
Abstract (max 300-500 words)
Poverty is not simply lack of financial resources; a broader more effective definition should
naturally involve a multitude of social-economics factors. The mere income distribution of a
country in fact results in ignoring social dynamics able to describe and predict poverty. This project
focuses on this multidimensional nature of poverty measurement taking into account factors
which can be used as indicator of wealth and social inequality. Such factors are usually categorized
into categorical variables and the interest is the underlying system of multivariate interdependences.
Marginal Models, as proposed by Bergsma and Rudas, 2002, appears to have the potential to
describe dependences/independences more precisely than traditional multivariate tools such as
logistic regression.
Moreover, we will focus on Chain Graphical Models (CGM), i.e. chain-graph-based models
recently proposed in the literature, which in addition offer a visual interpretation of the problem by
means of a graph where variables are represented as vertices and inter-relationships are represented
by the presence/absence of edges, as detailed for instance in Lauritzen (1996). CGMs are able to
represent complex structures of dependence/independence, taking advantage of the possibility of
grouping the variables into components. This allows for easily representing the variables as "purely
explicative", "purely response" and "intervening".
With the help of this statistical tools we will test all possible graphical models to
The main objective is about to find the suitable model(s) which best represents the real situation.
The selected model(s) & its graph will allow for analysing in a multivariate context whether and
how poverty is affected by the selected set of risk factors.
Introduction and motivation of visit
The main aim of my participation in this project is to deep my understanding of the poverty
phenomena within the social and economic reality. Another important goal is to improve my
knowledge about statistical models which I deepened in my PhD thesis (Supervisor: Prof. Colombi)
and in a successive paper (in collaboration with my Post-doctoral Supervisor: Prf. Mecatti), now
accepted for publication.
Scientific objectives of visit
My main expectation would be to clearly show the advantages of a multivariate approach to poverty
study. I believe that a non-standard recent statistical tool such as Chain Graphical Models would
add value to the analysis both for its advanced methodology and its results as clear to understand
and simple to be used even by non-technical users.
Reasons for choosing research infrastructure and datasets/surveys/...
BIGSSS offers the EU-SILC data and the support to understand the data.
2
Activities during your visit (research, training, events, ...)
The activities during my visit at BIGSSS entailed conducting research and analysis by using R
software, on the EU-SILC, 2011, improving the results obtained in my previous visiting to TARKI
with the same project.
Method and set-up of research
Datasets EU-SILC was managed with R statistical softwere with the package “hmmm”.
Project achievements during visit (and possible difficulties encountered)
Deep study of the link between social and wehalt situation through chain graphical model.
Preliminary project results and conclusions
Preliminary projects results are about to come.
Outcomes and future studies
The outcome of my visit at TARKI will be peer-reviewed and potentially published in a scientific
journal.
References







Bergsma, W. P., Rudas, T. (2002) Marginal models for categorical data. Annals of
Statistics, 30, 140-159.
Drton, M.: Discrete chain graph models. Bernoulli 15(3), 736–753 (2009)
Nicolussi, F.: Marginal parameterizations for conditional independence models and
graphical models for categorical data, PhD Thesis (2013)
Nicolussi, F., Colombi, R.: Graphical model of type II: a smooth subclass. Working
paper n. 224, http://boa.unimib.it/handle/10281/51950 (2014)
Rudas, T., Bergsma, W., Nemeth, R. (2010) Marginal log-linear parameterization of
conditional independence models. Biometrika, 97(4), 1006-1012.
R Core Team (2014). R: A language and environment for statistical computing. R
Foundation for
Statistical Computing, Vienna, Austria. URL http://www.Rproject.org/.
Colombi, R., Giordano, S., Cazzaro, M. (2014). hmmm: An R Package for Hierarchical
Multinomial Marginal Models. Journal of Statistical Software, 59(11), 1-25. URL
http://www.jstatsoft.org/v59/i11/.
3