Empirical Bayes model selection and analysis of microarray data

                    Matt Wand, UNSW
		    
Abstract

An emerging area of research interest for statisticians has been
in the analysis of data from DNA microarrays.  DNA microarrays
allow the simultaneous measurement of expression for thousands
of genes in tissue samples.  We consider
analysis of data from DNA microarray experiments where
scientific questions of interest can be framed in terms of comparison
of a collection of linear models for each of the genes.
We discuss ways in which we can use hierarchical and empirical
Bayes methods for model selection to borrow strength across
genes for making inferences and to deal with questions of
multiple comparisons and model uncertainty.  Differences between
our methodology and existing approaches to the same problem
will be discussed:  in particular, our methodology doesn't require
prespecification of a prior probability for differential
expression and we take a model selection rather than hypothesis
testing approach so that comparison of non-nested models for each
gene can be undertaken.  Application of the methods is considered
for a series of experiments intended to study the contribution
of genetic variation to control of gene transcription.  This
is joint work with Eva Chan, Chris Cotsapas, William Dunsmuir,
Michael Kirk, Peter Little, Rohan Williams and Matt Wand.

** This work is joint with David Nott