|
Activity Number:
|
93
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 12, 2002 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing*
|
| Abstract - #301255 |
|
Title:
|
A Comparison of Methods for Managing Type I Errors when Testing for Gene Expression Changes
|
|
Author(s):
|
Dan Nettleton*+
|
|
Affiliation(s):
|
Iowa State University
|
|
Address:
|
124 Snedecor Hall, Ames, Iowa, 50011-1210, USA
|
|
Keywords:
|
microarray ; false discovery rate ; bootstrap
|
|
Abstract:
|
The goal of many gene expression experiments is to determine which of several thousand genes change expression in response to one or more treatments. I will present results from two experiments, each involving the examination of over 8000 genes, in treated and control tissues. Each gene is tested for evidence of expression change using linear model techniques, coupled with bootstrapping of residuals to determine a p-value. I will discuss three methods for using these p-values to prepare a list of genes that are believed to have changed expression in response to treatment. Selecting genes with p-values less than some significance level (e.g., 0.05) is one option that will lead to many type I errors. The use of a resampling-based method that provides approximate strong control of the probability of one or more type I errors is a second option that should lead to few type I errors (but probably many type II errors). A method controlling the false discovery rate is presented as a third option that lies somewhere between the other two methods with regard to balance between type I and type II errors. Experimental design aspects crucial for these analyses will be emphasized.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |
