[BioC] edgeR: glmLRT test

James W. MacDonald jmacdon at uw.edu
Fri Mar 22 15:17:18 CET 2013


Hi KJ,

The short answer is that it doesn't matter, contingent upon you having 
set up the experiment correctly. Have you tried constructing the design 
matrix each way?

 > phenotype <- factor(rep(1:2, each=8))
 > area <- factor(rep(1:2, times = 8))
 > colnames(model.matrix(~area+phenotype))
[1] "(Intercept)" "area2"       "phenotype2"
 > colnames(model.matrix(~phenotype+area))
[1] "(Intercept)" "phenotype2"  "area2"

The interpretation of each coefficient will be the same in both instances.

Best,

Jim



On 3/22/2013 9:57 AM, KJ Lim wrote:
> Dear edgeR community,
>
> Good day.
>
> I have sets of RNA-Seq data of 2 phenotype (HighS, LowS) plants with area
> TZ&  SW (control).  I used exactTest to study which genes that are
> differential expressed in area TZ compare to SW of each phenotype.
>
> I would like to know do the phenotypes have effect on the area, in this
> case I should use glmLRT test. Before fit into the GLM, I defined my design
> matrix as:
>
>    ~phenotype+area
>
> Could someone please advice me regarding the design matrix? Should I define
> my design matrix as  ~area+phenotype ?
>
> Thank you very much for your time and help.
>
> Have a nice weekend.
>
> Best regards,
> KJ Lim
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

-- 
James W. MacDonald, M.S.
Biostatistician
University of Washington
Environmental and Occupational Health Sciences
4225 Roosevelt Way NE, # 100
Seattle WA 98105-6099



More information about the Bioconductor mailing list