Bayesian and Frequentist Handling of Multiple Testing M.J. Bayarri and
J.O. Berger
U. Valencia, Duke U. and SAMSI Abstract
It has been
frequently asserted that Bayesian methods for multiple testing
ignore the multiplicity issue; also that FDR can be interpreted in a
Bayesian fashion. In this talk we address these issues while reviewing
recent and popular methods to address multiple testing . We consider
both Bayesian and Frequentist methods, as well as eclectic ones that
seem to combine both approaches; in particular, we demonstrate how
Bayesian analysis handles multiplicities, and we look at methods
based on some type of False Discover Rate error. The main goal is to
gain understanding of these methods from both a Bayesian perspective
and a decision-theoretic perspective. For instance, a key question is
whether FDR can arise naturally in a decision setting. This talk
summarizes a variety of issues about multiple testing discussed during
the SAMSI Summer Program on Multiplicity and Replicability in
Scientific Studies.
|