[BioC] Limma and p-adjust: 'global' on non-orthogonal contrasts?

[Paolo Innocenti]

Dear Gordon and list,

I have a question about p-value adjustment for multiple testing across 
contrasts.

Let's say we have three groups, as in the Limma user guide, section 8.6, 
and this contrasts are closely related one another, so we want to apply 
method="global" to decideTests().

The code in section 8.6:

 > f <- factor(targets$Target, levels=c("RNA1","RNA2","RNA3"))
 > design <- model.matrix(~0+f)
 > colnames(design) <- c("RNA1","RNA2","RNA3")
 > fit <- lmFit(eset, design)
 > contrast.matrix <- makeContrasts(RNA2-RNA1, RNA3-RNA2, RNA3-RNA1,
+                                  levels=design)
 > fit2 <- contrasts.fit(fit, contrast.matrix)
 > fit2 <- eBayes(fit2)

Here we have 3 contrasts on 3 levels, so only the first two are 
orthogonal (or at least so it says in "The R book"). Is it justified 
here to run

 >decideTests(fit2, method="global")

or having non totally independent contrasts makes it impossible to 
correct the p-values "globally"?

Best,
paolo

-- 
Paolo Innocenti
Department of Animal Ecology, EBC
Uppsala University
Norbyvägen 18D
75236 Uppsala, Sweden