Without hacking into the code behind LIMMA, is there a way to perform partial F-tests for groups of variables outside of creating complicated contrast matrices? I have an experiment involving fixed time points where the researcher would like to test these two hypotheses: H0_1: Is any one of the early time points different from the other early time points; and H0_2: Is any one of the late time points different from the other late time points. It should be obvious that this is best attacked using a partial F, but how to implement this in LIMMA?? Thanks!! Dave H PS I'm in digest mode and will reply when I receive the digest. Alternatively cc me on your reply to the list. -- David A. Henderson, Ph.D. Assistant Professor of Animal Genetics, CALS Assistant Professor of Biostatistics, AZCOPH Member, The Bio5 Institute http://www.bio5.org 231 Shantz Building Department of Animal Sciences The University of Arizona Tucson, AZ 85721 Phone: (520)626-8903 Fax: (520)621-9435 DNADave@U.Arizona.Edu http://dnadave.cals.arizona.edu/

> Date: Tue, 08 Mar 2005 09:15:49 -0700 > From: David Henderson <dnadave@u.arizona.edu> > Subject: [BioC] Partial F-test using LIMMA > To: bioconductor@stat.math.ethz.ch > > Without hacking into the code behind LIMMA, is there a way to perform > partial F-tests for groups of variables outside of creating complicated > contrast matrices? I have an experiment involving fixed time points > where the researcher would like to test these two hypotheses: H0_1: Is > any one of the early time points different from the other early time > points; and H0_2: Is any one of the late time points different from the > other late time points. It should be obvious that this is best attacked > using a partial F, but how to implement this in LIMMA?? There is no way to compute F tests in limma other than by specifying contrast matrices. Feel free to explain a syntax which would be simpler. Gordon > Thanks!! > > Dave H > > PS I'm in digest mode and will reply when I receive the digest. > Alternatively cc me on your reply to the list. > -- > David A. Henderson, Ph.D. > Assistant Professor of Animal Genetics, CALS > Assistant Professor of Biostatistics, AZCOPH > Member, The Bio5 Institute http://www.bio5.org > 231 Shantz Building > Department of Animal Sciences > The University of Arizona > Tucson, AZ 85721 > Phone: (520)626-8903 > Fax: (520)621-9435 > DNADave@U.Arizona.Edu > http://dnadave.cals.arizona.edu/

Hello -- I'm working together with a statistical programmer here on a package that facilitates model assessment and comparison in the context where the same model was fit to lots of responses (=genes/probes), possibly using LIMMA, possibly not. Partial F tests for comparing nested models are certainly included in that. If you want to contact me off-list, I'd be happy to share what we have so far. Jenny