[BioC] limma and block effect

[Gordon Smyth]

>Date: Mon, 15 Aug 2005 15:30:34 -0700 (PDT)
>From: Wenbin Liu <wnbnl at yahoo.com>
>Subject: [BioC] limma and block effect
>To: bioconductor at stat.math.ethz.ch
>
>Dear limma users,
>
>I have an experiment designed as follows:
>
>Samples were taken from 3 individuals(S1, S2 and S3),
>and for the sample from each individual, there are
>four treatments (A,B, C, and D). Each sample was
>hybridized to an Affymetrix expression array.
>Therefore, the individuals serve as blocks.
>
>Will the following code work?
>
>treat <- as.factor(rep(c("A", "B", "C", "D"), 3))
>block <- rep(1:3, rep(4,3))
>design <- model.matrix(~ -1 + treat)

You cannot omit estimating the within-block correction. You must include
the duplicateCorrelation() step here, as in the example in the User's Guide.

>fit1 <- lmFit(dat, design=design, block=block)
># where dat is an ExprSet.
>
>In addition, what if I have a missing chip (11 chips
>instead of the balanced 12 ones)?

Makes no difference.

>  Will argument
>method='gls' do the job?

Not necessarly. Please follow the example in the User's Guide.

Gordon

>Any comment or hint will be greatly appreciated!
>
>Wenbin