Qing Li
 

Department of Mathematics, Washington University
 

Title: A two-way model for the normalization and inference of microarrays

 

Abstract: In two sample microarray experiments, the chief goal is to figure out the differentially expressed genes. To achieve this goal, one necessary step is to build a normalization model to remove the bias caused by the experiment system. Although people usually carry out the normalization and inference in two separate steps, some are trying to integrate them into one single analysis. In this talk, we will cover some tentative work in the integrated analysis, especially the inclusion of heteroscedasticity and the corresponding "small variance" problem. Some shrinkage estimators and the idea of empirical null distribution are adopted to alleviate the small variance problem.