[Forum SIS] DEC - Statistics Seminar -October 20th

[Raffaella Piccarreta]

Dear all, 


We are glad to announce the next: 

DEC - Statistics Seminar 

Thursday, October 20th 


Bocconi University, 
Room 3-E4-SR03 
Via Rontgen 1 - 3rd floor 

Time: 12:30 pm 




Mark Steel 
(University of Warwick) 

On choosing mixture components via non-local priors 

Abstract: 
Choosing the number of components remains a central but elusive challenge in mixture models. Traditional model selection criteria can fail to enforce parsimony or result in poorly separated components of limited practical use. Non-local priors (NLPs) are a family of distributions that encourage parsimony by enforcing a separation between the models under consideration. We formalize NLPs in the context of mixtures and show how they lead to extra parsimony and well-separated components that have non-negligible weight, hence interpretable as distinct subpopulations. We derive tractable expressions and suggest default prior settings aimed at detecting multi-modal densities. We also give a theoretical characterization of the sparsity induced by NLPs and propose easily implementable algorithms to obtain the integrated likelihood and parameter estimates. Although the framework is generic we fully develop the multivariate Normal mixture case based on a novel family of exchangeable moment priors. The proposal is illustrated using simulated and real data sets. Our results show a serious lack of sensitivity of the Bayesian information criterion (BIC) and insufficient parsimony of local prior and shrinkage counterparts to our formulation, which strikes a reasonable balance between power and parsimony. 



The DEC statistics seminars schedule is available at http://www.unibocconi.eu/statseminar 

Kind regards, 
Raffaella Piccarreta and Isadora Antoniano 

NOTE: 


If you are a guest and you do not have a Bocconi ID Card to access to the Bocconi Buildings, please confirm your participation by sending an email to simona.gagino a unibocconi.it 
In addition, we suggest to print, fill out and send the attached form (to give your consent to personal data handling) via email to the same address. In this way, the check-in procedure at the entrance will be faster and easier. 


*************************************************** 
Raffaella Piccarreta 
Associate Professor - Statistics 
DEC - Department of Decision Sciences, 
Universita' L.Bocconi, via Guglielmo Röntgen 1 (3rd floor, room D1-09). 20136, Milano. 
email: piccarreta a unibocconi.it 
tel. +39-02-58365659 / fax +39-02-58365630 

Web page (ENGLISH VERSION): http://faculty.unibocconi.eu/raffaellapiccarreta/ 
(ITALIAN VERSION): http://faculty.unibocconi.it/raffaellapiccarreta/ 
*************************************************** 
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20161014/3e583249/attachment-0001.html>
-------------- parte successiva --------------
Un allegato non testuale è stato rimosso....
Nome:        ModuloVisitorPass&Privacy.pdf
Tipo:        application/pdf
Dimensione:  56070 bytes
Descrizione: non disponibile
URL:         <http://www.stat.unipg.it/pipermail/sis/attachments/20161014/3e583249/attachment-0002.pdf>
-------------- parte successiva --------------
Un allegato non testuale è stato rimosso....
Nome:        M.Steel.20.10.16 .pdf
Tipo:        application/pdf
Dimensione:  24854 bytes
Descrizione: non disponibile
URL:         <http://www.stat.unipg.it/pipermail/sis/attachments/20161014/3e583249/attachment-0003.pdf>