In genome-scale RNAi screens, the primary objective is to select siRNAs with desired effect sizes, which relies on the comparison of gene effects in multiple different groups. Thus statistical methods for group comparisons play a critical role in data analysis in RNAi screens. A major statistical method for group comparison is contrast analysis. Traditionally, a contrast is a linear combination of group means in which the coefficients sum to zero. A typical contrast analysis is the significance testing of whether a contrast is zero. However, there are many issues with such contrast analysis. In fact, issues with the significance testing of a simple contrast (i.e., testing no mean difference between two groups) have incurred continuous calls for a critical reexamination of the common use of null hypothesis significance testing (NHST) in behavioral and social science, which has even led some researchers to advocate that the use of significance tests be banned in research. The major issues with traditional contrast analysis are discussed in Section 7.1. We face similar issues when we apply traditional methods for group comparison to analyze data from genome-scale RNAi screens.

Recently, a new method of contrast analysis was proposed to address issues in traditional contrast. This core of this new method is the concept of using a contrast variable, defined as a linear combination of random variables (with each variable representing random values in a group), instead of group means.