Dear colleague,

It is a pleasure to announce that our next RSS Highland Local Group meeting is a joint meeting between our RSS Highlands Local Group and the University of St. Andrews held in St. Andrews on the afternoon (2pm) of Wednesday 27th November. The meeting, which will be held in lecture theatre C of the Mathematics Building of the University of St. Andrews (North Haugh), will include two talks with refreshments mid-way offering an opportunity for chat.

The first talk is provided by Professor Dave Woods (University of Southampton) who has expertise in experimental design selection and assessment and will be talking about the decision-theoretic approach to Bayesian design with a focus on a new criterion for design selection.  An abstract for this talk which is titled 'Decision-theoretic design of experiments for model discrimination' is provided below.

The second talk is provided by Dr. Daniel Mortlock (Imperial College London) who has expertise in the problems of inference about the real world when made from incomplete or imperfect data. The talk will describe some schemes for formalising heuristic approaches to model testing within a Bayesian framework and will not assume a prior knowledge of astronomy or cosmology; this talk will provide useful and refreshing perspective to the biological or medical context of many of our meetings. An absract for this talk which is titled 'Bayesian Model Comparison in Astronomy and Cosmology' is provided below. 

The full details of the meeting with abstracts are provided below.

With my best wishes,

malcolm (Hall)


Joint RSS St Andrews meeting St Andrews lecture theatre C, Maths building, North Haugh) :
Wednesday 27th November 2013
 
2.00pm -2.05pm
Welcome
 
2.05pm – 2.55pm
Dave Woods (University of Southampton, UK)
Decision-theoretic design of experiments for model discrimination
 
2.55pm – 3.30pm
Coffee Break
 
3.30pm -4.20pm
Daniel Mortlock (Imperial College, London, UK)
Bayesian Model Comparison in Astronomy and Cosmology
 
http://www.mcs.st-andrews.ac.uk/StatsSeminars/index.shtml
 
Abstracts:
 
Dave Woods (University of Southampton, UK)
Decision-theoretic design of experiments for model discrimination
The design of any experiment is implicitly Bayesian, with prior knowledge being used informally to aid decisions such as which factors to vary and the choice of plausible causal relationships between the factors and measured responses. Bayesian methods allow uncertainty in these decisions to be incorporated into design selection through prior distributions that encapsulate information available from scientific knowledge or previous experimentation. Further, a design may be explicitly tailored to the aim of the experiment through a decision-theoretic approach with an appropriate loss function. 
This talk will review the decision-theoretic approach to Bayesian design and then focus on a new criterion for design selection when the aim of the experiment is discrimination between rival statistical models. Motivated by an experiment from materials science, we consider the problem of early stage screening experimentation to choose an appropriate linear model, potentially including interactions, to describe the dependence of a response on a set of factors. We adopt an expected loss for model selection which is a weighted sum of posterior model probabilities and introduce the Penalised Model Discrepancy (PMD) criterion for design selection. 
The use of this criterion is explored through a variety of issues pertinent to screening experiments, including the choice of initial and follow-up designs and the robustness of design performance to prior information. Designs from the PMD criterion are compared with those from existing approaches through examples. Further issues, such as reducing the computational burden of the method for experiments with a large number of contending models, will be addressed if time allows. Some directions of current and future research will also be discussed.


Daniel Mortlock (Imperial College, London, UK)
Bayesian Model Comparison in Astronomy and Cosmology
 Bayesian inference provides a self-consistent method of model comparison, provided that i) there are at least two models under consideration and ii) all the models in question have fully-specified and proper parameter priors.  Unfortunately, these requirements are not always satisfied in real world problems.  This is a particular difficulty in astronomy/cosmology: despite the existence of exquisitely-characterised measurements and quantitative physical models (i.e., sufficient to compute a believable likelihood), these models generally have parameters without well-motivated priors, making completely rigorous model comparison a formal impossibility.  Still, huge advances have been made in cosmology in particular in the last few decades, implying that model comparison (and testing) is possible in practise even without fully specified priors.  I review the above issues (without assuming any knowledge of astronomy or cosmology) and describe some schemes for formalising such heuristic approaches to model testing within a Bayesian framework.