[SEMINARIO] AIC difference for ergodic diffusion processes fromdiscrete observations

[stefano iacus <stefano.iacus@unimi.it>]

Il giorno 6 ottobre 2005, alle ore 14:30, il prof. Masayuki Uchida,  
Faculty of Mathematics
Kyushu University, Fukuoka, Japan terra' un seminario dal titolo:


"AIC difference for ergodic diffusion processes from discrete  
observations"

Presso l'aula seminari del Dipartimento di Scienze Economiche,  
Aziendali e Statistiche, sita al 2 piano in Via Conservatorio 7,  
Milano. Tutti gli interessati sono invitati a partecipare.


Abstract:
Under the situation where parametric models contain a true model,  
Akaike's information criterion (AIC) is a functional tool for  
evaluating the statistical models estimated by the maximum likelihood  
estimators.　AIC is a criterion constructed by means of the maximum  
log likelihood function and the dimension of the parameter space.　 
For that reason, it seems easy to get AIC even for discretely  
observed multi-dimensional diffusion processes.　However, there is a  
serious problem to obtain AIC for discretely observed diffusion  
processes.　A transition density of a diffusion process does not  
generally have an explicit form.　Therefore, in most cases, the  
likelihood function is not obtained explicitly. In this talk, instead  
of obtaining AIC for discretely observed ergodic multi-dimensional  
diffusion processes, we treat AIC difference, which means the  
difference between Akaike's information criteria of two statistical  
models.　Since it is possible to obtain the AIC difference by means  
of a contrast function based on a locally Gaussian approximation, the  
proposed AIC difference can be a useful tool for model selection of  
diffusion processes.



----------------------------------------------------------------
Stefano M. Iacus
Dipartimento di Scienze Economiche, Aziendali e Statistiche
Universita' di Milano
Via Conservatorio, 7
I-20123 Milano
Tel: 02 50321 461
Fax: 02 50321 505



=============================================================================
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/
.
=============================================================================