|
Activity Number:
|
379
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
ENAR
|
| Abstract - #304570 |
|
Title:
|
Statistical Modeling for Oligonucleotide Arrays Using PCA with Likelihood Approach
|
|
Author(s):
|
Mehdi Maadooliat*+ and Jianhua Hu and Jianhua Huang
|
|
Companies:
|
Texas A&M University and The University of Texas M.D. Anderson Cancer Center and Texas A&M University
|
|
Address:
|
Deaprtment of Statistics, College Station, TX, 77840,
|
|
Keywords:
|
Oligonucleotide array ; SVD modes ; Principal Component ; Box-Cox transformation
|
|
Abstract:
|
Using appropriate model with adequate number of parameters could be always challenging. Here we are considering a simple PCA problem, based on the first mode and we would like to use a parametric transformation like Box-Cox to obtain normally distributed errors. Our motivation is oligonucleotide arrays which are widely used type of expression microarray data. Li and Wong have been proposed a popular model for obtaining the "expression index." Later on Hu and et al. showed the SVD interpretation for this model and they used grid search with entropy criteria to find an appropriate transformation for the arrays. Here the goal is to develop an automatic procedure to find the expression index, relying more on the statistical theories. Also we propose some level of assessment which shows that sometimes it's helpful to involve more SVD modes in gene expression models.
|
