Hi, I am analyze two color experiment with a common reference. It is a two factor experiment with the following targets : Cy3 Cy5 IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control IE Control IE Control EE Control EL Control IE Control IL Control EE Control IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control EL Control IE Control IL Control EE Control EL Control IE Control IE Control EE Control EL Control where the first I/E is thetreatment factor and the second E/L is the time factor. after preparing a design matrix with the argument ref="Control" I tried to fit above coefficients. >From limma userguide chapter 9.3 Common Reference Designs it is suggested to fit the coefficients on RGList object: fit <- lmFit(RG, design) when I tried this I got an error message Error in array(x, c(length(x), 1), if (!is.null(names(x))) list(names(x), : dim<- : dims [product 362880] do not match the length of object [8] It seems to me that it is more reasonable to fit the coefficients on normalized dat and indeed when I trried to fit it on MAList object all went well. Is it it a mistake to fit the coefficient on MAList object? Ron

Hi Ron When using limma I always fit the coefficients to a (normalised) MAList object. However, the documentation for lmFit() states: Fit linear model for each gene given a series of arrays lmFit(object,design=NULL,ndups=1,spacing=1,block=NULL,correlation=0.75 ,w eights=NULL,method="ls",...) object: object of class 'numeric', 'matrix', 'MAList', 'marrayNorm', 'exprSet' or 'PLMset' containing log-ratios or log-values of expression for a series of microarrays So lmFit() is basically a generic function which fits a linear model for each gene based on *some data*, and so passing it an RGList is valid if the RGList values are the data you want to use to fit linear models. Note however that the docs do insist on either log-ratios or log- values. It's definitely not a mistake to use an MAList object though, so go right ahead. Mick -----Original Message----- From: Ron Ophir [mailto:ron.ophir@weizmann.ac.il] Sent: 16 November 2004 09:16 To: bioconductor@stat.math.ethz.ch Subject: [BioC] limma design matrix with common reference Hi, I am analyze two color experiment with a common reference. It is a two factor experiment with the following targets : Cy3 Cy5 IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control IE Control IE Control EE Control EL Control IE Control IL Control EE Control IE Control IL Control EE Control EL Control IE Control IL Control EE Control EL Control EL Control IE Control IL Control EE Control EL Control IE Control IE Control EE Control EL Control where the first I/E is thetreatment factor and the second E/L is the time factor. after preparing a design matrix with the argument ref="Control" I tried to fit above coefficients. >From limma userguide chapter 9.3 Common Reference Designs it is suggested to fit the coefficients on RGList object: fit <- lmFit(RG, design) when I tried this I got an error message Error in array(x, c(length(x), 1), if (!is.null(names(x))) list(names(x), : dim<- : dims [product 362880] do not match the length of object [8] It seems to me that it is more reasonable to fit the coefficients on normalized dat and indeed when I trried to fit it on MAList object all went well. Is it it a mistake to fit the coefficient on MAList object? Ron _______________________________________________ Bioconductor mailing list Bioconductor@stat.math.ethz.ch https://stat.ethz.ch/mailman/listinfo/bioconductor

> Date: Tue, 16 Nov 2004 11:16:15 +0200 > From: "Ron Ophir" <ron.ophir@weizmann.ac.il> > Subject: [BioC] limma design matrix with common reference > To: <bioconductor@stat.math.ethz.ch> > Message-ID: <s199e19e.016@wisemail.weizmann.ac.il> > Content-Type: text/plain; charset=US-ASCII > >>From limma userguide chapter 9.3 Common Reference Designs it is > suggested to fit the coefficients on RGList object: > fit <- lmFit(RG, design) > when I tried this I got an error message > Error in array(x, c(length(x), 1), if (!is.null(names(x))) > list(names(x), : > dim<- : dims [product 362880] do not match the length of object > [8] > It seems to me that it is more reasonable to fit the coefficients on > normalized dat and indeed when I trried to fit it on MAList object all > went well. Sorry, this is a typo in the User's Guide. It doesn't make sense to fit a linear model to an RGList object, and trying to do so will lead to the error above. You are quite correct, the linear model should be fit to the normalized MAList object. Gordon > Ron