|
Activity Number:
|
164
- Random and Mixed Effect Models
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 31, 2017 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #323908
|
View Presentation
|
|
Title:
|
A Modified Mixed Model Approach to the Large Scale Multiple Testing Problem
|
|
Author(s):
|
John Grego* and Paramita Chakraborty and Chong Ma and James Lynch
|
|
Companies:
|
Univ of South Carolina and Department of Statistics, University of South Carolina and University of South Carolina Columbia and Department of Statistics, University of South Carolina
|
|
Keywords:
|
empirical Bayes ;
exponential tilting
|
|
Abstract:
|
In big data situations, such as microarray analysis, one source of lack of reproducibility is the number of false positives/discoveries that occur. One method uses a mixture contamination model to identify real significant (contaminant) cases based on the false discovery rate for each case. We propose a modification in which the model and screening criteria are adjusted by conditioning test results based on a general class of statistics. We use cross-validation to estimate the mixture distribution with one portion of the data and screen cases with the remaining portion. The cross-validation process allows us to explore biological networks for significant genes.
|
Authors who are presenting talks have a * after their name.
