Dear David, I don't know what the problem is here, but it is unlikely to have anything to do with missing values in your data. 1. Does you data object contain any real data? Try summary(exprs(myEset)) 2. What version of R and limma are you using? (Every person who posts a question to this list should be giving this information routinely.) If you upgrade to the latest limma (from CRAN) does the problem go away? 3. What packages do you have loaded in your session, and what other functions do you have in your current workspace? Does the problem go away if you run your analysis in a "clean" R session? Best wishes Gorodn > Date: Wed, 8 Feb 2006 17:19:13 +0100 (CET) > From: kfbargad at ehu.es > Subject: [BioC] Limma missing values > To: bioconductor at stat.math.ethz.ch > > Dear users, > > I am trying to use limma for analysis of Q-PCR data. > > I have a matrix of log2 intensities for 96genes X 20 samples that I > have imported using read.table. Some of the cells (not many) have > missing values because the PCR did not work for that gene on that > sample and I left them blank. >> myEset > Expression Set (exprSet) with > 96 genes > 19 samples > phenoData object with 1 variables and 19 cases > varLabels > cov1: read from file > > I read in the help vignette that lmFit can handle missing values, but > when I try it, I get the following error: > > treatments = factor(c(0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,1),labels = c > ("BD","C")) >> design = model.matrix(~0+treatments) >> fit <- lmFit(exprs(myEset),design) > Error in lm.fit(design, t(M)) : NA/NaN/Inf in foreign function call > (arg 4) > In addition: Warning message: > NAs introduced by coercion > > I checked the mailing archives and found a few emails saying that > limma can handle missing values. I also tried setting na.rm=TRUE > > Any ideas as to how to proceed? > > I also learned elsewhere that it is better to use a non-parametric > test for PCR data. Any opinions about this? > > Regards, > > David