University of Edinburgh
School of Mathematics and BioSS
Date: Friday 4th Nov, 16:10pm Location: JCMB 5327
Speaker: Prof Chris Holmes, Department of Statistics, University of Oxford
Title: Bayesian nonparametric hypothesis tests
Abstract: Hypothesis testing, such as two-sample tests and tests for
independence, are important components of biomedical data analysis. We
discuss recent advances in the use of Bayesian nonparametric (BNP)
models, such as Polya Trees and Dirichlet Processes, for this task. One
advantage of BNP methods is that they provide an explicit model based
probability measure for the null hypothesis Pr(H0 | data), as well as for the
alternative, Pr(H1 | data) = 1 - Pr(H0 | data), without making parametric
assumptions on the form of the likelihood function. This allows one to
incorporate prior information and control for multiple testing, while
naturally extending to tests for contrasts where we wish to detect for an
effect under one condition but not under another. This latter task is
problematic for non-Bayesian methods as one of the tests has a higher
dimension null than the alternative, but is readily accommodated in the
Bayesian approach. We present some examples from biomedical genomics.
--
There will be tea and coffee after the seminar in the Mathematics Common
Room.
This seminar is a part of Maxwell Institute seminar series.
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|