Hi everyone, I'm trying to test the effect of different medical parameters (fators) on my data with limma but I get a troncated design matrix. I've two factors 'UNOS' and 'Etiology' (with 3 and 2 levels respectively) listed below: Target.txt file looks like: sample UNOS Etiology s1 1 CMD s2 1 CAD s3 1B CMD s4 1B CAD S5 2 CMD S6 2 CAD ... I reading it as a phenoData file. I would like to know the different effects so I create a design matrix: >design<-model.matrix(~-1+Etiology*UNOS,data=pData(target)) >design EtiologyCAD EtiologyCMD UNOS1B UNOS2 EtiologyCAD:UNOS1B EtiologyCAD:UNOS2 EtiologyCMD:UNOS1B EtiologyCMD:UNOS2 but UNOS1 factor level is missing in the design matrix, Am I doing something wrong ? Thanks, Nolwenn ******************************************** Nolwenn Le Meur INSERM U533 Facult? de m?decine 1, rue Gaston Veil 44035 Nantes Cedex 1 France Tel: (+33)-2-40-41-29-86 (office) (+33)-2-40-41-28-44 (secretary) Fax: (+33)-2-40-41-29-50 mail: nolwenn.lemeur@nantes.inserm.fr

Actually, I think will result in the same problem. What you probably want is to use a contrast matrix to get what you need. All of the coefficients are with respect to Unos1. So you need to make a contrast matrix to get effects of interest (if they are different than the contrast with unos1). You can change the factor levels if you want some other level as the base group. In what contrasts are you interested? On Oct 6, 2004, at 8:41 AM, Nolwenn Le Meur wrote: > Sean, > Thanks for your answer. So if I understand instead of: > > sample UNOS Etiology > s1 1 CMD > s2 1 CAD > s3 1B CMD > s4 1B CAD > S5 2 CMD > S6 2 CAD > ... > > I should write: > > sample UNOS1 UNOS1B UNOS2 Etiology > s1 1 0 0 CMD > s2 1 0 0 CAD > s3 0 1 0 CMD > s4 0 1 0 CAD > S5 0 0 1 CMD > S6 0 0 1 CAD > ... > > Nolwenn > ******************************************** > Nolwenn Le Meur > INSERM U533 > Facult? de m?decine > 1, rue Gaston Veil > 44035 Nantes Cedex 1 > France > > Tel: (+33)-2-40-41-29-86 (office) > (+33)-2-40-41-28-44 (secretary) > Fax: (+33)-2-40-41-29-50 > mail: nolwenn.lemeur@nantes.inserm.fr > ******************************************** > > -----Message d'origine----- > De : Sean Davis [mailto:sdavis2@mail.nih.gov] > Envoy? : mercredi 6 octobre 2004 14:30 > ? : Nolwenn Le Meur > Objet : Re: [BioC] limma troncated design matrix > > > This is correct. You can't specify columns for every factor level or > your design matrix will be singular. > > Sean > > On Oct 6, 2004, at 6:30 AM, Nolwenn Le Meur wrote: > >> Hi everyone, >> >> I'm trying to test the effect of different medical parameters (fators) >> on my >> data with limma but I get a troncated design matrix. >> I've two factors 'UNOS' and 'Etiology' (with 3 and 2 levels >> respectively) >> listed below: >> >> Target.txt file looks like: >> sample UNOS Etiology >> s1 1 CMD >> s2 1 CAD >> s3 1B CMD >> s4 1B CAD >> S5 2 CMD >> S6 2 CAD >> ... >> >> I reading it as a phenoData file. >> >> I would like to know the different effects so I create a design >> matrix: >> >>> design<-model.matrix(~-1+Etiology*UNOS,data=pData(target)) >>> design >> EtiologyCAD EtiologyCMD UNOS1B UNOS2 EtiologyCAD:UNOS1B >> EtiologyCAD:UNOS2 >> EtiologyCMD:UNOS1B EtiologyCMD:UNOS2 >> >> but UNOS1 factor level is missing in the design matrix, Am I doing >> something >> wrong ? >> >> Thanks, >> >> Nolwenn >> >> >> ******************************************** >> Nolwenn Le Meur >> INSERM U533 >> Facult? de m?decine >> 1, rue Gaston Veil >> 44035 Nantes Cedex 1 >> France >> >> Tel: (+33)-2-40-41-29-86 (office) >> (+33)-2-40-41-28-44 (secretary) >> Fax: (+33)-2-40-41-29-50 >> mail: nolwenn.lemeur@nantes.inserm.fr >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor