Dear list, I am trying to figure out what normalization method would suit my needs best. To determine this, I thought about plotting mean expression value versus variance, both calculated across each probeset for one chip. Calculating the mean over a probeset is easy, as I can use the expresso method for it: expresso(data, bg.correct=FALSE, normalize=FALSE, pmcorrect.method="pmonly", summary.method="avgdiff") Not as easy is the calculation of the variance over each probeset. I wrote my own method for it, but it takes ages... My data comes from the MOE 430 2 Affymetrix GeneChip with 45101 probesets. I got 6 chips in total. Is anybody able to think of a faster way to compute the variance? Below is the code I was using. Thanks for your help! Sue --------------- getprobes <- function(genelist,data){ as.vector(t(pm(data,genelist))) } ##### ### INPUT1: exprSet ### INPUT2: raw Data (AffyBatch) meanvar <- function(exp,data){ split.screen(c(3,2)) # 3 rows, 2 columns list<-geneNames(exp) list<-as.array(list) for(j in 1:6){ r <- apply(list,1,getprobes,data[,j]) v <- lapply(r,var) screen(j) plot(exprs(exp)[,j],v,pch=".",main=paste("mean vs variance for chip ",deparse(j))) } } ______________________________________________________________________ ___ Mit der Gruppen-SMS von WEB.DE FreeMail k?nnen Sie eine SMS an alle