|
Activity Number:
|
504
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305787 |
|
Title:
|
Estimating the Positive False Nondiscovery Rate and False Nondiscovery Rate in Multiple Hypothesis Testing
|
|
Author(s):
|
Changzheng Chen*+ and Burt Holland
|
|
Companies:
|
K & L Consulting Services Inc. and Temple University
|
|
Address:
|
3 Birkdale Court, Westampton, NJ, 08060,
|
|
Keywords:
|
multiple testing ; FDR ; FNR ; PFNR ; n-value ; type II error
|
|
Abstract:
|
We present new results concerning Type II error concepts termed False Non-discovery rate (FNR) and positive False Non-discovery rate (pFNR) appropriate for multiple testing using FDR or pFDR for Type I error control. The application of these concepts depends on the nature of the scientific investigation under study and the characteristics of the family of hypotheses being tested, particularly its proportion of false hypotheses. We specify scenarios in which FNR and pFNR are more appropriately used than other Type II error concepts. For a fixed rejection region we provide point and bootstrap interval estimators of the FNR and pFNR that are applicable for large families. Conditions under which these estimators are conservative are discussed. We introduce the n-value, a pFNR analogue of observed Type II error. The new methodology is applied to gene expression data.
|
