|Multiple Comparison Procedures for Gene Expression Data
Igor V. Melnykov, PhD
Department of Mathematics and Physics
Colorado State University - Pueblo
Monday, March 20, 2006
B101 Engineering Building
Genetic microarrays present powerful means for the analysis of genetic information. In a microarray experiment, genes (or fragments of genes) are printed on a slide to form an array consisting of hundreds of spots each representing a particular gene. The relative abundance of each gene sequence in different DNA samples can then be determined. The analysis of data resulting from this type of experiment presents interesting statistical challenges. Thus, when making simultaneous inference on all genes printed on the array, one must make multiplicity adjustments to ensure the desired significance level. This approach is realized, for example, in a well-known stepwise procedure by Holm. We are looking at the improvement of testing power due to considering sharper bounds in the case of exchangeable variables. The use of measures related to false discovery rate (FDR) can be viewed as an alternative to the approaches controlling the probability of at least one false positive. The FDR is defined as the expected proportion of false rejections. We consider some convergence properties of the positive FDR.