F-test in Microarray Experiments
Authors:Kristen Cunanan, Atousa Karimi, Duy Ngo, Calvin Pham, Suzette Puente, Emily Ramos
Mentor:Gulhan Bourget, Associate Professor of Statistics, California State University Long Beach
Microarray experiments measure thousands of genes at once to identify differentially expressed genes under different conditions. For example, comparisons can be made between genes from the same tissue (cancerous and non-cancerous liver) or different tissues (brain and skin). In a complex disease, researchers are interested in discovering disease causing genes in humans. It has been discovered that humans have approximately 30,000 genes. Microarray experiments can study, for example, all those 30,000 genes at once. If all genes are studied, a microarray experiment produces 60,000 gene expressions from healthy and diseased people. After the collection of data, the next step is to analyze gene expressions. The first step in data analysis is to study if there is a difference in the mean expressions of 30,000 genes from both groups (healthy versus diseased). A common approach is to use the classical $F-$test. One of the assumption of this test is that genes are independent. However, in nature, many genes can act together; hence, the assumption of independency is violated. Nevertheless, many researchers are still using this test. In this research, we investigate through Monte Carlo Simulation study to see if the $F$-test can still be used. Our findings show that the $F$-test can be considered in analysis if the dependency among genes is small to mild.