>Hi, >I have a set of affymetrix data which includes three groups. I would >like once to run supervised analysis by performing all three pairwise >comarisons and once unsupersied by clustering samples. The question is: >Should I run the clustring on the normalized observed data (one that >comaes out of RMA) or on coefficients after fitting in order that the >input for the two type of analysis would be comparable? >Having the following experimental design: > A B C >Array1 1 0 0 >Array2 1 0 0 >Array3 0 1 0 >Array4 0 1 0 >Array5 0 0 1 >Array6 0 0 1 >like in second design of chapter 13 "Two Groups: Affymetrix", is the >coefficients after lmFit(data,design) are the same as the data itself? after running fit<-lmFit(data,design) fit$coefficients are actually the avrage values of (Array1,Array2) (Array3,Array4) (Array5,Array6). Thus for each gene you'l get a vector of size 3 which is exactly the averages of the groups you have been described in the design matrix. Therefore, If you would like to cluster the groups based on averages use fit$coeff matrix and if you want to cluster the sample based on the replicate values use the exprs(AffyBatch) matrix. >Thanks, >Ron