Asbtract: Principal components have been used a lot with microarray data to exhibit unwanted variation, and to some extent to remove such variation. Such efforts are also called normalization. In this talk I will review some of these methods, and related work, and propose a simple but apparently novel variant which works a lot of the time, though not always. Part of my story will be a discussion of how one tells whether one is helping or hurting the analysis by adjusting.
Joint work with Johann Gagnon-Bartsch