Hi, In limma, what is the best way to automatically generate a contrast matrix with all possible pairs? For example, if I have 4 groups, is there a way to generate the matrix with 6 columns each corresponding to a pair-wise comparison, other than having to specify the 6 expressions explicitly when calling makeContrasts? I tried to first create a vector with those expressions and use that in makeContrasts: > pair<-c() > ind<-1 > for (i in 2:numg){ + for (j in i:numg) { + pair[ind] <- paste(groupname[j], "-", groupname[(j-i+1)]) + ind<- ind+1}} > pair [1] "GroupB - GroupA" "GroupC - GroupB" "GroupC - GroupA" > makeContrasts(pair, levels=design) pair GroupA "GroupB - GroupA" GroupB "GroupC - GroupB" GroupC "GroupC - GroupA" A related problem: I noticed that expressions have to be separated by comma. However, even when I tried: > makeContrasts(pair[1], levels=design) pair[1] GroupA "GroupB - GroupA" GroupB "GroupB - GroupA" GroupC "GroupB - GroupA" Compare to the result by doing: > pair[1] [1] "GroupB - GroupA" > makeContrasts("GroupB - GroupA", levels=design) GroupB - GroupA GroupA -1 GroupB 1 GroupC 0 The help page says: "...: expressions, or character strings which can be parsed to expressions" But my string doesn't seem to be converted. Any suggestions? Will it be more convenient to use if the ... argument is a vector of expressions or something similar? Thank you, Yiwen NIH/CIT

It is true that makeContrasts() doesn't accept character-valued variables. E.g., > makeContrasts("A-B",levels=c("A","B")) > makeContrasts(A-B,levels=c("A","B")) will both work but > x <- "A-B" > makeContrasts(x,levels=c("A","B")) will not. I will have a think about how this might be fixed. To compute a design matrix to make all possible pairwise comparisons, it is easier to do this directly. Here is a little function which does the job: > design.pairs <- function(levels) { n <- length(levels) design <- matrix(0,n,choose(n,2)) rownames(design) <- levels colnames(design) <- 1:choose(n,2) k <- 0 for (i in 1:(n-1)) for (j in (i+1):n) { k <- k+1 design[i,k] <- 1 design[j,k] <- -1 colnames(design)[k] <- paste(levels[i],"-",levels[j],sep="") } design } > design.pairs(c("A","B","C","D")) A-B A-C A-D B-C B-D C-D A 1 1 1 0 0 0 B -1 0 0 1 1 0 C 0 -1 0 -1 0 1 D 0 0 -1 0 -1 -1 Gordon > Date: Mon, 6 Jun 2005 18:22:11 -0400 > From: "He, Yiwen (NIH/CIT)" <heyiwen@mail.nih.gov> > Subject: [BioC] makeContrasts in limma > To: "'bioconductor@stat.math.ethz.ch'" > <bioconductor@stat.math.ethz.ch> > > Hi, > In limma, what is the best way to automatically generate a contrast matrix > with all possible pairs? For example, if I have 4 groups, is there a way to > generate the matrix with 6 columns each corresponding to a pair-wise > comparison, other than having to specify the 6 expressions explicitly when > calling makeContrasts? > > I tried to first create a vector with those expressions and use that in > makeContrasts: > >> pair<-c() >> ind<-1 >> for (i in 2:numg){ > + for (j in i:numg) { > + pair[ind] <- paste(groupname[j], "-", groupname[(j-i+1)]) > + ind<- ind+1}} >> pair > [1] "GroupB - GroupA" "GroupC - GroupB" "GroupC - GroupA" > >> makeContrasts(pair, levels=design) > pair > GroupA "GroupB - GroupA" > GroupB "GroupC - GroupB" > GroupC "GroupC - GroupA" > > A related problem: I noticed that expressions have to be separated by comma. > However, even when I tried: > >> makeContrasts(pair[1], levels=design) > pair[1] > GroupA "GroupB - GroupA" > GroupB "GroupB - GroupA" > GroupC "GroupB - GroupA" > > Compare to the result by doing: >> pair[1] > [1] "GroupB - GroupA" >> makeContrasts("GroupB - GroupA", levels=design) > GroupB - GroupA > GroupA -1 > GroupB 1 > GroupC 0 > > The help page says: > "...: expressions, or character strings which can be parsed to expressions" > > But my string doesn't seem to be converted. > > Any suggestions? Will it be more convenient to use if the ... argument is a > vector of expressions or something similar? > > Thank you, Yiwen > NIH/CIT