Dear all, I am working with limma package for the analysis of differential expression in agilend double color chips. My data come from a common reference experiment and the design matrix is the following: WT MU 1 0 -1 0 0 1 0 -1 1 0 -1 0 0 1 0 -1 Since rows 1 and 2 are technical dye-swap replicates (as rows 3 and 4, 5 and 6, 7 and 8), I am trying to use duplicateCorrelation with the block argument as follows: rep<-(1,1,2,2,3,3,4,4). But there's something that I don't understand. 1. if I give the following design matrix (instead of the one written above) (1,-1,1,-1,1,-1,1-1), the result in corfit$consensus changes (but not very much) why? 2. if I give as design matix the abs(.) of the matrix above, the result in confirt$consensus changes very much. why? I thought that duplicateCorrelation needs the design matrix to be able to distinguish between WT and MU. I thought it was't necessary to say which chip is dye-swap (it should come from the logratio values?). And why the result changes if I give the vector in 1.? What exactly duplicateCorrelation calculates? Thak you very much! Roberta Sirovich -- ----------------------------- Dr. Roberta Sirovich Department of Mathematics University of Torino Via Carlo Alberto 10 10123 Torino, Italy tel: +39-011-6702850 fax: +39-011-6702878