[BioC] limma/multtest comparison question

Gordon Smyth smyth at wehi.edu.au
Fri Jun 18 04:52:18 CEST 2004

topTable() in limma always does adjustment for multiple testing. The 
default is the very conservative 'Holm" method. You probably want to choose 
something less conservative.


At 12:40 PM 18/06/2004, Sean Davis wrote:
>Dear all,
>I have a limma/multtest question.  I have performed what I think is the 
>equivalent of a t-test (of course, moderated) in limma below.  Compare 
>that with the output from mt.maxT in the multtest package.  Why are there 
>such large discrepancies (ie., what am I doing wrong), particularly in 
>p-values?  The data (abinorms3) are single-channel log2 intensities.  The 
>commands and output are below.  We have the same samples hybridized on 
>two-color arrays, also analyzed with limma, with p-values in mt.maxT and 
>limma that agree with the mt.maxT result from below, so I think there are 
>differentially-expressed genes.

More information about the Bioconductor mailing list