A Multiple Test with Application to Gene Microarray Data

• Activity Number: 334
• Type: Contributed
• Date/Time: Wednesday, August 6, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Sin-Ho Jung*+
• Companies: ACOSOG
• Keywords: adjusted p value ; multistep procedure ; Bonferroni test ; t-test ; two-sided test ; Wilcoxon rank sum test

Abstract

Microarray technology is rapidly emerging for genome-wide screening of differentially expressed genes between critical subtypes of human diseases. Traditional statistical testing approaches, such as two-sample t-test or Wilcoxon test are frequently used for evaluating statistical significance of such differential expressions. However, these approaches require adjustment for a serious multiple comparison problem. Some recent studies suggested the use of Bonferroni-adjusted tests or multistep permutation tests to circumvent this difficulty. However, the Bonferroni tests are often found too conservative and the multistep permutation tests computationally expensive to implement and difficult to interpret their results. We propose a one-step multiple testing procedure to derive an exact critical value under the desired level of the family-wise error rate. Our approach is based on a permutation method to derive the null distribution that preserves the original structure of all genes' dependency.