Statistics seminar
Liverpool University Statistics Division
Time: 2pm, Tuesday September 21st
Venue: Room 2.11, Maths & Oceanography Building, University of Liverpool
Speaker: Alex Karagrigoriou, University of Cyprus
Title: On asymptotic properties of model identification procedures
Abstract: Model selection is the area of statistics that examines the
problem of selecting an appropriate model for a given set of observations.
The main issues in model selection are consistency and asymptotic efficiency
which are both associated with the goodness of fit of the selected model.
The goodness of the selection depends on the object of the analysis.
Consistency is the main concern in case the true model is known as correctly
as possible and asymptotic efficiency is of great importance if the object
of the analysis is a selection that yields good inference.
In this talk we attempt a literature survey of the above asymptotic properties.
First, we provide some of the most popular of the model identification
procedures such as AIC (Akaike, 1973), FPE (Akaike, 1969), $FPE_a$ (Bhansali
and Downham, 1977), BIC (Schwarz, 1978), $\varphi$ (Hannan and Quinn, 1980),
FIC (Wei, 1992). Then, some consistency results are presented (Shibata, 1976;
Hannan and Quinn, 1979; Hannan, 1980; Bhansali, 1986; Knight, 1989; Wei,
1992). Finally the concept of asymptotic efficiency is thoroughly discussed
and results for 1-step ahead prediction are presented from the time as well as
the frequency domain point of view (Shibata, 1980, Karagrigoriou, 1995, 1997).
Finally, the multiple prediction case is discussed and results by Bhansali
(1996) and Karagrigoriou and Vonta (1999) are presented.
Other forthcoming seminars:
Wednesday 13 October
Anthony Ledford (University of Surrey)
Wednesday 20 October
Gesine Reinert (University of Cambridge)
Full seminar programme available via http://www.liv.ac.uk/maths/SOR/
Damian Clancy
-----------------
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|