Microarray Data Analysis: Comparing Two Population Means

Scientists employing microarray profiling technology to compare sample sets generate data for a large number of endpoints. Assuming the experimental design minimized sources of bias, and the analytical technology was reliable, precise, and accurate, how does the experimentalist determine which endpoints are meaningfully different between the groups? Comparison of two population means for individual analysis measurements is the most common statistical problem associated with microarray data analysis. This chapter focuses on the hands-on procedures using SAS software to describe how to choose statistical methods to find the statistically significantly different endpoints between two groups of data generated from reverse phase protein microarrays. The four methods outlined are: (a) two-sample t -test, (b) Wilcoxon rank sum test, (c) one-sample t -test, and (d) Wilcoxon signed rank test. Two sample t -test is used for two independently normally distributed groups. One-sample t -test is used for a normally distributed difference of paired observations. Wilcoxon rank sum test is considered a nonparametric version of the two-sample t -test, and Wilcoxon signed rank test is considered a nonparametric version of the one-sample t -test.