[BioC] What is limma saying about my data?

[michael watson (IAH-C)]

Hi

Briefly, I am using R 2.0.1 and limma 1.8.6.  I have a double dye-flip
experiment with only four arrays, and my targets and design matrix is
so:

> targets
  SlideName                                         File       Cy3
Cy5
1  TB_43_97           Untreated Cy3/TB_43_97/results.gpr Untreated
H202
2 TB_43_106          Untreated Cy3/TB_43_106/results.gpr Untreated
H202
3  TB_43_99           Untreated Cy5/TB_43_99/results.gpr      H202
Untreated
4 TB_43_101 Untreated Cy5/TB_43_101/results 750vs630.gpr      H202
Untreated

> design
[1]  1  1 -1 -1

After running:

> fit <- lmFit(MA, design)
> fit <- eBayes(fit)
> topTable(fit, adjust="fdr")

I find that EVERY single one of my p-values is 0.9999963, apart from the
last 10 or so which are NA (as all the data for those spots is NA).
Looking at the top gene, it has values of -1.9, -2.9, 2, 1.5, which when
flipped the right way and given to t.test show a significant result (for
what it's worth).

So, why are all my p-values 0.9999963?  Have I done something wrong (it
is just after Christmas)?  Is my data seriously screwed, or is my
limma/R seriously screwed?  Has anyone seen this before?

Cheers
Mick