[Statlist] RESEARCH SEMINAR IN STATISTICS - UNIVERSITY OF GENEVA

Eva Cantoni Ev@@C@nton| @end|ng |rom un|ge@ch
Mon Mar 24 12:41:27 CET 2014


RESEARCH SEMINAR IN STATISTICS - UNIVERSITY OF GENEVA

Organisers : .
E. Cantoni - E. Ronchetti - S. Sperlich - M-P. Victoria-Feser

Friday March 28th, 2014
at 11h15
Room M 5220, Uni Mail (40, bd du Pont-d'Arve)


Gael MARTIN
Monash University, Melbourne

"Approximate Bayesian Computation in State Space Models"

Abstract:
Exploiting the likelihood-free techniques of approximate Bayesian 
computation, a new approach to inference in state space models is 
proposed. Building on the concept of the asymptotic sufficiency of the 
maximum likelihood estimator (MLE), the MLE of the parameters of an 
auxiliary, or approximating, model forms the basis for the summary 
statistic in an ABC matching criterion. The augmented unscented Kalman 
filter (AUKF) is used for this purpose, with the speed and simplicity of 
the AUKF calculations being critical to the success of the ABC scheme. 
Consideration is given to the very concept of sufficiency in the state 
space context, and to the effectiveness of ad-hoc summary measures in 
this setting. The application of the proposed method in multiple 
parameter settings is an important consideration, with a separate 
treatment of scalar (or lower-dimensional blocks of) parameters 
proposed, based on integrated likelihood methods. Particular emphasis is 
given to the situation where the state variable (and possibly the 
observed) is driven by a continuous time process, in which case exact 
Bayesian inference is typically infeasible as a result of intractable 
transitions. A stochastic volatility model for financial returns, based 
on a square root specification for volatility, is used for illustration.

joint work with

Brendan P.M. McCabe (Management School, University of Liverpool, U.K.),
Christian P. Robert (University of Dauphine and CREST, Paris) and
Ole Maneesoonthorn (Melbourne Business School, University of Melbourne, 
Australia).


Visit the website: http://www.stat-center.unige.ch/ResSem.html

-- 
Prof. Eva Cantoni
Research Center for Statistics and
      Geneva School of Economics and Management
University of Geneva, Bd du Pont d'Arve 40, CH-1211 Genève 4
http://www.unige.ch/ses/dsec/staff/faculty/Cantoni-Eva.html




More information about the Statlist mailing list