[BioC] Limma parameters and FDR

Gordon Smyth smyth at wehi.edu.au
Tue Dec 9 00:29:59 MET 2003


I'm not sure that I understand the question. The default proportion in 
eBayes is not the false discovery rate, as you probably already know. You 
are free to select as many genes as you like your topTable from limma, 
going as far down into the higher false discovery rates as you like. Just 
change the 'number' parameter to ask for more genes. The Benjamini and 
Hochberg idea is that, wherever you stop, you have controlled the FDR at 
that level. (Note however that this makes assumptions which are not true 
for microarray data.) There is no 0.05 default in limma.


At 02:50 AM 9/12/2003, you wrote:
>As an additional but related question. Is it possible to change the
>proportion of falsely rejected hypotheses given using the Benjamini and
>Hocberg fdr procedures in multtest or the p.adjust/topTable functions in
>I suspect that 0.05 is also too conservative
>-----Original Message-----
>From: Stephen Henderson
>To: 'bioconductor at stat.math.ethz.ch'
>Sent: 08/12/03 13:49
>Subject: [BioC] Limma parameters
>  I've been examining the eBayes function in Limma using some affymetrix
>we have. The default proportion setting: 0.01 I think is probably wrong
>the comparison of tissue groups I have.
>Is there a rational way to estimate/guesstimate the correct proportion?
>isn't that the point anyway?

More information about the Bioconductor mailing list