|
Activity Number:
|
135
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305441 |
|
Title:
|
Model-Based Correlations: a Tool for Revealing Interactions in Microarray Data
|
|
Author(s):
|
Beatrix Jones*+ and Marie Fitch
|
|
Companies:
|
Massey University and Massey University
|
|
Address:
|
Institute of Information and Mathematical Sciences, Auckland, 0000, New Zealand
|
|
Keywords:
|
covariance selection ; factor analysis ; graphical models ; microarray data
|
|
Abstract:
|
Correlation is a basic and popular tool to look for potential interactions among many continuous variables, as occur in a microarray experiment. As well as thresholded raw correlations, one can use model-based correlations (e.g. derived from Bayesian covariance selection or factor analysis models). Interactions also can be examined using the inverse covariance, which represents the conditional independence structure. Do these more computationally intensive methods offer advantages---better or complementary information about interactions? Using simulated data and microarray data from yeast, we examine the abilities of these methods to reveal and represent the known interactions present.
|
