Hi, I'm sorry if this this question is probably basic and statistically related. I am analyzing 24 affy slides taken from GEO, 12 slides for each of the two conditions. According to the limma vignetetts chapter 8.5, I can have kinds of two design matrix. What are the differences? So I can either do "treatment-contrasts parameterization": design <- cbind(NORMAL=1, NORMALvsCANCER=c(rep(0, times=12), rep(1, times=12))) or linear contrasts: design <- cbind(NORMAL=c(rep(1, times=12), rep(0, times=12)), CANCER=c(rep(0, times=12), rep(1, times=12))) I am not sure which to choose, I currently uses the second method. Timothy [[alternative HTML version deleted]]