Search
Question: reference argument in limma factorial design
0
gravatar for Gordon Smyth
9.7 years ago by
Gordon Smyth35k
Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia
Gordon Smyth35k wrote:
> Date: Thu, 15 Jan 2009 16:12:08 -0500 > From: Wei Xu <xuwei3 at="" msu.edu=""> > Subject: [BioC] reference argument in limma factorial design > To: bioconductor at stat.math.ethz.ch > Content-Type: text/plain > > Hi, everyone, > > I am working on a dual-channel cDNA microarray project which includes 32 > slides as a loop design. It is a 2x2x2x2 factorial experimental design that > combines 16 treatments. When I was trying to use modelMatrix to design the > matrix, it always asked me to input a reference. However, in my experiment, > I don't really have one as a reference. I tried to assign a random treatment > as reference but that way I couldn't make contrasts of the reference with > the other treatments. Yes, you can make contrasts between the reference and other treatments, in fact that is what the coefficients now mean. Please refer to the examples on common reference designs in the limma User's Guide (for example Section 7.3), or past advice given on this list. Best wisehs Gordon > How should I set this argument? Is there a different > way that allows me not to assign a reference? Thanks. > > Wei > > -- > Wei Xu > Aquatic Animal Health Laboratory > Pathobiology and Diagnostic Investigation > College of Veterinary Medicine > Michigan State University > East Lansing, MI, 48824 > Tel: 1-517-432-8258
ADD COMMENTlink modified 9.7 years ago by Naomi Altman6.0k • written 9.7 years ago by Gordon Smyth35k
0
gravatar for Naomi Altman
9.7 years ago by
Naomi Altman6.0k
Naomi Altman6.0k wrote:
I think that Wei Xu is referring to the fact that one ANOVA is not an identifiable model without a constraint - hence one either has to construct the model matrix with an indicator variable, or with the factor effects. In limma, this does not matter, because limma does the overall F-test and then allows you to construct as many contrasts as are reasonable. I always use single channel analysis for loop designs, which makes it easier to set up the main effects and interactions. This was a kind of "statsy" comment on my part, but I hope that those who have taken a course including factorial designs can follow. --Naomi At 10:33 PM 1/16/2009, Gordon K Smyth wrote: >>Date: Thu, 15 Jan 2009 16:12:08 -0500 >>From: Wei Xu <xuwei3 at="" msu.edu=""> >>Subject: [BioC] reference argument in limma factorial design >>To: bioconductor at stat.math.ethz.ch >>Content-Type: text/plain >> >>Hi, everyone, >> >>I am working on a dual-channel cDNA microarray project which includes 32 >>slides as a loop design. It is a 2x2x2x2 factorial experimental design that >>combines 16 treatments. When I was trying to use modelMatrix to design the >>matrix, it always asked me to input a reference. However, in my experiment, >>I don't really have one as a reference. I tried to assign a random treatment >>as reference but that way I couldn't make contrasts of the reference with >>the other treatments. > >Yes, you can make contrasts between the reference and other >treatments, in fact that is what the coefficients now mean. Please >refer to the examples on common reference designs in the limma >User's Guide (for example Section 7.3), or past advice given on this list. > >Best wisehs >Gordon > >>How should I set this argument? Is there a different >>way that allows me not to assign a reference? Thanks. >> >>Wei >> >>-- >>Wei Xu >>Aquatic Animal Health Laboratory >>Pathobiology and Diagnostic Investigation >>College of Veterinary Medicine >>Michigan State University >>East Lansing, MI, 48824 >>Tel: 1-517-432-8258 > >_______________________________________________ >Bioconductor mailing list >Bioconductor at stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor >Search the archives: >http://news.gmane.org/gmane.science.biology.informatics.conductor Naomi S. Altman 814-865-3791 (voice) Associate Professor Dept. of Statistics 814-863-7114 (fax) Penn State University 814-865-1348 (Statistics) University Park, PA 16802-2111
ADD COMMENTlink written 9.7 years ago by Naomi Altman6.0k
Thanks, Gordon and Naomi. One of the examples in limma manual talks about 2x2 factorial experiments of a two-channel array. It estimates the interaction in the makeContrasts. > cont.matrix <- makeContrasts( + WT11.MT11=P11mt-P11wt, + WT21.MT21=P21mt-P21wt, + WT11.WT21=P21wt-P11wt, + MT11.MT21=P21mt-P11mt, + Int=(P21mt-P11mt)-(P21wt-P11wt), + levels=design) Since I got more factors (2x2x2x2), can I make the interactions with more factors (i.e. two-way, three-way, and four-way interactions)? Wei On Fri, Jan 16, 2009 at 11:10 PM, Naomi Altman <naomi@stat.psu.edu> wrote: > I think that Wei Xu is referring to the fact that one ANOVA is not an > identifiable model without a constraint - hence one either has to construct > the model matrix with an indicator variable, or with the factor effects. > > In limma, this does not matter, because limma does the overall F-test and > then allows you to construct as many contrasts as are reasonable. > > I always use single channel analysis for loop designs, which makes it > easier to set up the main effects and interactions. > > This was a kind of "statsy" comment on my part, but I hope that those who > have taken a course including factorial designs can follow. > > --Naomi > > > > At 10:33 PM 1/16/2009, Gordon K Smyth wrote: > > Date: Thu, 15 Jan 2009 16:12:08 -0500 >>> From: Wei Xu <xuwei3@msu.edu> >>> Subject: [BioC] reference argument in limma factorial design >>> To: bioconductor@stat.math.ethz.ch >>> Content-Type: text/plain >>> >>> Hi, everyone, >>> >>> I am working on a dual-channel cDNA microarray project which includes 32 >>> slides as a loop design. It is a 2x2x2x2 factorial experimental design >>> that >>> combines 16 treatments. When I was trying to use modelMatrix to design >>> the >>> matrix, it always asked me to input a reference. However, in my >>> experiment, >>> I don't really have one as a reference. I tried to assign a random >>> treatment >>> as reference but that way I couldn't make contrasts of the reference with >>> the other treatments. >>> >> >> Yes, you can make contrasts between the reference and other treatments, in >> fact that is what the coefficients now mean. Please refer to the examples >> on common reference designs in the limma User's Guide (for example Section >> 7.3), or past advice given on this list. >> >> Best wisehs >> Gordon >> >> How should I set this argument? Is there a different >>> way that allows me not to assign a reference? Thanks. >>> >>> Wei >>> >>> -- >>> Wei Xu >>> Aquatic Animal Health Laboratory >>> Pathobiology and Diagnostic Investigation >>> College of Veterinary Medicine >>> Michigan State University >>> East Lansing, MI, 48824 >>> Tel: 1-517-432-8258 >>> >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >> > > Naomi S. Altman 814-865-3791 (voice) > Associate Professor > Dept. of Statistics 814-863-7114 (fax) > Penn State University 814-865-1348 (Statistics) > University Park, PA 16802-2111 > > > -- Wei Xu Aquatic Animal Health Laboratory Pathobiology and Diagnostic Investigation College of Veterinary Medicine Michigan State University East Lansing, MI, 48824 Tel: 1-517-432-8258 [[alternative HTML version deleted]]
ADD REPLYlink written 9.7 years ago by Wei Xu100
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 343 users visited in the last hour