ABSTRACT
An important and common task in microarray analysis is to detect differentially expressed genes by comparing the expression levels of thousands of genes in samples collected at two different conditions. The microarray data usually contain thousands of genes, but only a very limited number of samples on each gene. Thus, the microarray experiments present statisticians a big challenge with its “large p small n” paradigm (West et al., 2000). To solve this issue, many researches have been conducted from both the frequentist and the Bayesian framework, to propose powerful models for analyzing the microarray data, and efficient model selection algorithms for identifying the best subset of genes to be differentially expressed (DE). The detailed descriptions of the existing methods are available in several review articles and books, including Pan (2002), Kuo et al. (2008), Parmigiani et al. (2003), Speed (2003), Wit and McClure (2004), and Lee (2004).
