[Statlist] RESEARCH SEMINAR IN STATISTICS - UNIVERSITY OF GENEVA

[Eva Cantoni]

RESEARCH SEMINAR IN STATISTICS - UNIVERSITY OF GENEVA

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

Friday April 4th, 2014
at 11h15
Room M 5220, Uni Mail (40, bd du Pont-d'Arve)

Mark HANNAY
Université de Genève

"Confidence sets for models in the linear setting"

Abstract:
Model selection might, at first glance, seem clear. Out of a list of 
possible models we want to select the "best" or a subset of "best" 
models. This notion of "best" however is not well defined, since it 
obviously depends on the initial goals of the selection. Such goals 
might include : finding the smallest "true" model, finding the "best" 
model for prediction or even finding the model with minimal MSE of the 
estimate. Throughout, we will focus on prediction.
One common approach is to define the "best" model as the model with 
maximum mean expected log likelihood (MELL).
This is a useful definition, when trying to optimize prediction, since 
the MELL rates how well the estimate of a specific model would perform 
on new data.
Some criteria already exist to minimize the MELL, most notably the AIC 
and BIC criteria. Both these criteria are very useful in providing a 
single model. However alone they do not provide any account on the 
variability of the selection. Just as for estimation of parameters we 
would like to have something resembling confidence intervals for model 
selection.
By using the MELL, we propose confidence sets for models in the linear 
setting. These confidence sets will then provide us with a notion of 
variability in model selection with confidence levels.
Although we mainly focus on the linear case, these results can easily be 
extended to other types of models.


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