| The Human Genome Project is the most important reason for the surge of new technologies in the microarray area. These technologies facilitate the experimentation with a large number of genes simultaneously. These experiments allow, for example, to compare the genetic expressions between sick and healthy cells. One of the main goals in analysis of gene expression data from microarray experiments is the identification of genes differentially expressed. To accomplish this goal various methods are used. One of them is to use multiple testing establishing a null hypothesis of not association between the measure of gene expression and the response is tested for each gene. Due to the fact that in microarray experiments thousand of genes are considered, the same amount of tests must be considered. In this situation, the probability of error type I increases with the number of hypothesis. Thus, a p-value near zero can occur randomly due to the large number of hypothesis that are tested simultaneously and not because the gene is differentially expressed.;In this thesis, we study methods for multiple testing on data coming from microarray experiments. We compare theoretically and experimentally ten multiple hypothesis testing methods on six gene expression data sets, which are freely available in the internet. |
