IMS
Session Slot: 2:00-3:50 Sunday
Estimated Audience Size: 125-175
AudioVisual Request: Two Overheads
Session Title: Model Uncertainty
Theme Session: No
Applied Session: No
Session Organizer: Draper, David University of Bath, UK
Address: Department of Mathematical Sciences University of Bath Claverton Down, Bath BA2 7AY England
Phone: +44-1225-826222
Fax: +44-1225-826492
Email: d.draper@maths.bath.ac.uk
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 0 minutes First Speaker - 30 minutes Second Speaker - 30 minutes Third Speaker - 30 minutes Discussant - (none) Floor Discusion - 15 minutes
Session Chair: Draper, David University of Bath, UK
Address: Department of Mathematical Sciences University of Bath Claverton Down, Bath BA2 7AY England
Phone: +44-1225-826-222
Fax: +44-1225-826-492
Email: David Draper <d.draper@maths.bath.ac.uk>
1. A Bayesian decision-theoretic approach to model selection
Walker, Stephen, Imperial College, London
Address: Department of Mathematics Imperial College of Science, Technology, and Medicine Huxley Building 180 Queen's Gate London SW7 2BZ England
Phone: +44-171-594-8522
Fax: +44-171-594-5817
Email: S G Walker <s.walker@ic.ac.uk>
Abstract: We put the problem of model selection into a Bayesian decision-theoretic framework. Typically, the models under consideration consist of a finite number of parametric models. However, the decision space might also include a nonparametric model. We therefore consider the problem of deciding when to adopt a nonparametric approach in preference to "sticking" with a parametric model.
2. Bayesian model averaging
Madigan, David, University of Washington
Address: Department of Statistics, GN-22 University of Washington Seattle WA 98195-0001 USA
Phone: 206-543-4537
Fax: 206-685-7419
Email: David Madigan <madigan@stat.washington.edu
Abstract: Standard statistical practice ignores model uncertainty. Data analysts typically select a model from some class of models and then proceed as if the selected model generated the data. This approach ignores the uncertainty in model selection, leading to over-confident inferences.Bayesian model averaging (BMA) provides a coherent mechanism for accounting for this model uncertainty. Several methods for implementing BMA have recently emerged. We discuss these methods and present a number of examples. We emphasize applications involving latent variables, missing data, and causal inference, where between-model uncertainty can overwhelm within-model uncertainty.
3. Model-based inference for categorical survey data
Forster, Jon, University of Southhampton, UK
Address: Department of Mathematics University of Southampton Southampton SO17 1BJ England
Phone: +44-1703-595-130
Fax: +44-1703-595-147
Email: Jon Forster <jjf@maths.soton.ac.uk>
Abstract: For survey data, model uncertainty may relate both to the underlying "data model" and to models for "non-sampling errors" such as nonresponse bias. In this talk, I consider categorical data and show how different approaches may be used to account for such uncertainty in order to reliably estimate quantities of interest.