An Exploration of Aspects of Bayesian Multiple Testing
James Scott and James Berger
There has been increased interest of late in the Bayesian approach
to multiple testing (often called the multiple comparisons
problem), motivated by the need to analyze DNA microarray data in
which it is desired to learn which of potentially several thousand
genes are activated by a particular stimulus. We study the issue
of prior specification for such multiple tests; computation of key
posterior quantities; and useful ways to display these quantities.
A decision-theoretic approach is also considered.