|
Activity Number:
|
497
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309630 |
|
Title:
|
Bayesian Semiparametric Testing and Estimation for High-Dimensional Genetic Data
|
|
Author(s):
|
Richard MacLehose*+ and David B. Dunson
|
|
Companies:
|
National Institute of Environmental Health Sciences and National Institute of Environmental Health Sciences
|
|
Address:
|
PO 12233, Research Triangle Park, NC, 27709,
|
|
Keywords:
|
Dirichlet Process ; SNP ; Bayes
|
|
Abstract:
|
Genotyping technology has made analysis of large dimensional data commonplace, with the number of predictors typically exceeding the number of observations. Statistical methods are needed to test associations between a vast number of SNPs and a phenotypic outcome, while accounting for high levels of between SNP correlation. False discovery rates are often used in this setting to control the expected proportion of false positives, however they typically ignore the problem of false negatives. We adopt a decision theoretic Bayesian approach which penalizes false positives and false negatives. To reduce dimensionality, we focus on the clustering properties of the Dirichlet Process (DP). We adopt a "retrospective" approach to estimating DP models that is more robust to high dimensions. We discuss various procedures for hypothesis testing and compare our method to FDR-based methods.
|
