When I created the design matrices in the original Weaver example, I just created the design matrices myself using cbind(). You could however use modelMatrix() using the 'param' argument to specify what you want the coefficients to represent. Gordon On Wed, June 1, 2005 1:32 am, Wenbin Liu said: > Thanks, Gordon! In the expanded example, you have a > common RNA source, the Pool. What if one doesn't have > a common RNA source, such as when it is a loop design > with more than three arrays like the unexpanded Weaver > data? Can you please show me how to generate the > matrices? Thanks! > > > Wenbin > > --- Gordon K Smyth <smyth@wehi.edu.au> wrote: > >> I guess that you are using an older version of the >> software. The Weaver example was expanded and >> simplified last October, and full commands are now >> given to compute the design matrix, including >> use of modelMatrix() and makeContrasts(). >> >> Gordon >> >> ------ original message -------- >> [BioC] limma example Weaver Data >> Wenbin Liu wnbnl at yahoo.com >> Thu May 26 23:25:09 CEST 2005 >> >> Dear Gordon and limma users, >> >> I'm struggling to understand the design matrices in >> the Weaver Data in the Case Studies section of the >> limma User's Guide. >> >> In the single loop design, there are 4 arrays and 4 >> conditions (from two time points and two genotypes). >> I tried using modelMatrix() with the target from the >> example and assigning ref='P11WT'. I would like to >> check the effect 'TimeWt' which is the late vs. >> early >> for WT. Unfortunately, I have been unable to get >> Gordon's result (topTable result). >> >> Can someone show me how to get Gordon's design >> matrices (two of them in that example) by using the >> function modelMatrix() and probably plus >> makeContrasts()? >> >> Thanks in advance! >> >> Sincerely, >> Wenbin

The easiest way to handle these sort of direct designs (e.g., loop designs) in general is to choose one of the RNA sources to play the role of a reference source for the purposes of the design matrix. Have you read the section on "Direct Two-Color Designs" in the limma User's Guide? In limma 1.9.6, this is Section 9.4 of the guide. The design matrix given on page 22 could also have been created using design <- modelMatrix(targets, ref="CD4") Gordon At 01:32 AM 1/06/2005, Wenbin Liu wrote: >Thanks, Gordon! In the expanded example, you have a >common RNA source, the Pool. What if one doesn't have >a common RNA source, such as when it is a loop design >with more than three arrays like the unexpanded Weaver >data? Can you please show me how to generate the >matrices? Thanks! > >Wenbin