|
Activity Number:
|
609
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #305650 |
|
Title:
|
Multiple Imputation for Microarray Missing Data
|
|
Author(s):
|
Hui Xie*+ and Leann Myers and Steven Smith
|
|
Companies:
|
Tulane University and Tulane University and Pennington Biomedical Research Center
|
|
Address:
|
1440 Canal Street, Suite 2001, New Orleans, LA, 70112,
|
|
Keywords:
|
Multiple Imputation ; Gibbs samplers ; Wishart distribution ; missing data ; Microarray ; Data Augmentation
|
|
Abstract:
|
Microarray technology as a tool to analyze genome wide mRNA gene expression profiles has been dramatically developed in last two decades. Missing points in the microarray data sets are inevitable for a variety of technical and biological reasons. Imputation of the missing data to facilitate the complete data requirement for the statistical procedures is an appealing solution. In this study, we focus on multiple imputation via the MCMC approach implemented with the Gibbs sampler and data augmentation. Two experimental sampling designs are considered: pooling units and single units. Because of non-normal distribution of microarray raw data (log transformed) with heavy right-tail skew and finite boundaries, we propose to draw random samples from truncated multivariate normal distribution rather than the ordinary one, while the covariance matrix is assumed to follow the Wishart distribution.
|
