|
Activity Number:
|
131
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #300421 |
|
Title:
|
Compound False Positive Rate and Compound Power: An Alternative Approach to Multiple-Hypothesis Testing in Microarray Analysis
|
|
Author(s):
|
Xuejun Peng*+ and Arne C. Bathke and Constance L. Wood and Richard J. Kryscio and Arnold J. Stromberg
|
|
Companies:
|
The University of Kentucky Math Sciences and University of Kentucky and University of Kentucky and University of Kentucky and University of Kentucky
|
|
Address:
|
4070 Victoria Way Apt. 34, Lexington, KY, 40515-4692,
|
|
Keywords:
|
microarray ; multiple-hypothesis testing ; cFPR ; cPOW ; power ; sample size
|
|
Abstract:
|
The promising field of genomics has brought many new challenges to statistical inference research. Microarray data offer a new level of complexity of multiple-hypothesis testing adjustment. Classical procedures that control FWER have in general low power and prove to be of limited use in analyzing microarray data where only a small proportion of genes may actually be differentially expressed. The False Discovery Rate (FDR) control also has its limitations. In this paper, we first define compound False Positive Rate (cFPR) and Compound Power (cPOW) in multiple testing. Then we provide algorithms to estimate cFPR and cPOW using the empirical distribution of the p values. We also discuss how to estimate the sample size needed to control cFPR and cPOW to the desired level. The performance of different approaches in terms of cFPR and cPOW is compared using simulated as well as real microarray data. We show that two modified Sidak procedures strike better balance between Type I and Type II error rates and hence are more suitable for large scale hypothesis testing, such as in microarray data analysis.
|
