|
Activity Number:
|
421
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #310009 |
|
Title:
|
Bayesian Variogram Modeling in a Metric Space
|
|
Author(s):
|
Yan Zheng*+ and Cavan Reilly
|
|
Companies:
|
sanofi-aventis and The University of Minnesota
|
|
Address:
|
1212 Cornerstone Blvd, Downingtown, PA, 19335,
|
|
Keywords:
|
Bayesian inference ; gene expression ; Bessel functions ; correlation functions ; variograms ; Gibbs' sampler
|
|
Abstract:
|
We propose an approach to Bayesian variogram modeling for data that is referenced to a metric space and apply the method to the analysis of Affymetrix oligonucleotide arrays. Before fitting a parametric model to the correlation structure, we investigated the possibility of anisotropy. We mapped the oligonucleotide space to a metric space using multidimensional scaling method and the directional semivariograms are used to assess anisotropy. We investigate the impact of the sequence on the measured intensity and fit a 25-way ANOVA to the intensity with factors being the nucleotide in each of the 25 positions of the 25mer. Finally we computed the residuals from the model and fit Bessel mixture Bayesian variogram models for different number of Bessel mixture components on these residuals.
|
