Confronting Intractability in Statistical Inference
Research workshop: 16-19 April 2012, Bristol, UK
Aims and Objectives
Many problems currently faced by statistical scientists involve the mathematical or computational intractability of likelihoods and other objects which are central to data analysis. This intractability may arise from the sheer size of data sets, the requirement to work with physical models and computer simulations, or from the complexity of statistical models which have many unobserved elements and parameters of large, or possibly infinite dimension.
This workshop will bring together a group of world-leading researchers whose work addresses this intractability, in its various forms, with a diverse set of techniques.
The aim is to create an event at which experts highlight the sources of difficulty in their respective fields of specialisation, describe recent developments which treat these issues and thereby foster opportunities for transfer of ideas and methodology. Topics of interest include, but are not limited to:
- Inference in models on high and infinite dimensional spaces, via Monte Carlo methodology and deterministic approximations
- The applied mathematics/statistics interface, data assimilation and inference on function spaces
- Approximate Bayesian computation and composite likelihood methods
- Statistical treatment of computer experiments and complex dynamical models
This workshop is organised by Nick Whiteley, Christophe Andrieu and Mark Beaumont.