Hi, I have a question regarding setting up a two way anova with two factors: organ (5 levels) and treatment (2 levels). I am not sure quite how to set this up. I understand from the user guide in limma how to set up a 2 x 2. And I understand how to set up the interaction term by subtracting the differences, but having 5 levels has stumped me. Can you explain a way to set it up? I am using a Nimblegen array with approximatel 20,000 probes and I have 30 arrays in my columns for my eset (3 replicates of each treatment and organ). Here is what I was using in my script: trts = c("B.A","B.C","C.A","C.C","L.A","L.C","R.A","R.C","S.A","S.C") > f = gl(10,3,labels=trts) > design <- model.matrix(~0+f) > colnames(design) <- trts > fit <- lmFit(eset, design) > contrasts <- makeContrasts(OxT =(B.A-B.C)-(C.A-C.C)-(L.A-L.C)-(R.A-R.C)-(S.A-S.C),O = (B.A+B.C)-(C.A+C.C)-(L.A+L.C)-(R.A+R.C)-(S.A-S.C),T = (B.A+C.A+L.A+R.A+S.A)-(B.C+C.C+L.C+R.C+S.C),levels=design) > fit2 <- contrasts.fit(fit, contrasts) > fit2 <- eBayes(fit2) I don't think I did this correctly. Any suggestions? I also tried defining the factors with different levels, but I got an error: > levels(f) <- list(B=c(1:6), Cu=c(7:12), L=c(13:18), R=c(19:24), S=c(25:30)) > Organ <- factor(eset$Organ, levels=c("B","Cu","L","R","S")) > levels(f) <- list(A=c(1:3,7:9,13:15,19:21,25:27), C=c(4:6,10:12,16:18,22:24,28:30)) > Treatment <- factor(eset$Treatment, levels=c("A","C")) > contrasts(Organ) <- contr.sum(5) > contrasts(Treatment) <- contr.sum(2) > design <- model.matrix(~Organ*Treatment) Warning message: In is.na(rows) : is.na() applied to non-(list or vector) of type 'NULL' Thanks for your help in advance! Grant R. Cramer Professor Department of Biochemistry and Molecular Biology, MS 330 University of Nevada, Reno Reno, NV 89557 (775) 784-4204 cramer at unr.edu http://www.ag.unr.edu/cramer/

Dear Professor Cramer, Why not design <- model.matrix(~Organ*Treatment,data=pData(eset)) ? If that doesn't work, then you need to show us what pData(eset) contains. Also, the choice of design matrix does depend on what hypotheses you want to test, and you haven't said anything about that. Best wishes Gordon > Date: Fri, 14 Oct 2011 16:15:41 -0700 > From: Grant R Cramer <cramer at="" unr.edu=""> > To: "bioconductor at stat.math.ethz.ch" <bioconductor at="" stat.math.ethz.ch=""> > Subject: [BioC] two way ANOVA with limma > Message-ID: <fb9ac11a-598d-469e-8b75-2d0121c6257a at="" unr.edu=""> > Content-Type: text/plain; charset="us-ascii" > > Hi, > I have a question regarding setting up a two way anova with two factors: > organ (5 levels) and treatment (2 levels). I am not sure quite how to > set this up. I understand from the user guide in limma how to set up a 2 > x 2. And I understand how to set up the interaction term by subtracting > the differences, but having 5 levels has stumped me. Can you explain a > way to set it up? > I am using a Nimblegen array with approximatel 20,000 probes and I have > 30 arrays in my columns for my eset (3 replicates of each treatment and > organ). > > Here is what I was using in my script: > > trts = c("B.A","B.C","C.A","C.C","L.A","L.C","R.A","R.C","S.A","S.C") >> f = gl(10,3,labels=trts) >> design <- model.matrix(~0+f) >> colnames(design) <- trts >> fit <- lmFit(eset, design) >> contrasts <- makeContrasts(OxT =(B.A-B.C)-(C.A-C.C)-(L.A-L.C)-(R.A-R.C)-(S.A-S.C),O = (B.A+B.C)-(C.A+C.C)-(L.A+L.C)-(R.A+R.C)-(S.A-S.C),T = (B.A+C.A+L.A+R.A+S.A)-(B.C+C.C+L.C+R.C+S.C),levels=design) >> fit2 <- contrasts.fit(fit, contrasts) >> fit2 <- eBayes(fit2) > > I don't think I did this correctly. Any suggestions? > > I also tried defining the factors with different levels, but I got an error: > > >> levels(f) <- list(B=c(1:6), Cu=c(7:12), L=c(13:18), R=c(19:24), S=c(25:30)) >> Organ <- factor(eset$Organ, levels=c("B","Cu","L","R","S")) >> levels(f) <- list(A=c(1:3,7:9,13:15,19:21,25:27), C=c(4:6,10:12,16:18,22:24,28:30)) >> Treatment <- factor(eset$Treatment, levels=c("A","C")) >> contrasts(Organ) <- contr.sum(5) >> contrasts(Treatment) <- contr.sum(2) >> design <- model.matrix(~Organ*Treatment) > > Warning message: > In is.na(rows) : is.na() applied to non-(list or vector) of type 'NULL' > > Thanks for your help in advance! > > Grant R. Cramer > Professor > Department of Biochemistry and Molecular Biology, MS 330 > University of Nevada, Reno > Reno, NV 89557 > (775) 784-4204 > cramer at unr.edu > http://www.ag.unr.edu/cramer/ ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}