Hi, I have some questions about the chapter 9 in limma user's guide. Please forgive my lack of statistical knowlege. 1. For the example in section 9.4 technical replication II >design<-designMatrix(targets, ref="wt1") >fit<-lmFit(MA, design) >cont.matrix<-makeContrasts(MuvsWt=(mu1+mu2-wt2)/2, levels=design) >fit2<-contrasts.fit(fit,cont.matrix) This technique includes an effect for each biological replicate. How about technical replicates, does it treat technical replicates as biological replicates? Am I right that this technique doesn't consider effect for technical replicate? 2. For the dye swap example in section 9.3 technical replication I >design<-c(1,-1,1,-1) >corfit<-duplicateCorrelation(MA, design,ndups=1,block=c(1,1,2,2)) >fit<-lmFit(MA,design, block=c(1,1,2,2), correlation=corfit$consensus) This technique handles technical replicates by using the function duplicateCorrelation. Am I right it includes an effect for each technical replicates, but not an effect for each biological replicate? how is biological replicate's effect handled here? I couldn't find statistician who specialized in microarray in my area, could you explain it briefly or point me to some direction to study on it? I would appreciate any help. Ren [[alternative HTML version deleted]]

> This technique includes an effect for each biological replicate. How about > technical replicates, does it treat technical replicates as biological > replicates? Am I right that this technique doesn't consider effect for > technical replicate? Biological replicates may have systematic difference (e.g. different individuals), but technical replicates are purely replicates ( 2 samples from the same individual) which we generally don't think they have effect beside error. > This technique handles technical replicates by using the function > duplicateCorrelation. Am I right it includes an effect for each technical > replicates, but not an effect for each biological replicate? how is > biological replicate's effect handled here? No, actually this include an (random) effect for each biological replicate, no effect for technical replicates. block=c(1,1,2,2) means each each biological replicate form one block. Hopefully this will help. Fangxin -- Fangxin Hong, Ph.D. Bioinformatics Specialist Plant Biology Laboratory The Salk Institute 10010 N. Torrey Pines Rd. La Jolla, CA 92037 E-mail: fhong@salk.edu