[BioC] Fwd: SAFE with small sample sizes

[Ina Hoeschele]

----- Forwarded Message -----
From: "Ina Hoeschele" <inah at vbi.vt.edu>
To: bioconductor at stat.math.ethz.ch
Sent: Monday, October 25, 2010 10:34:22 AM
Subject: Re: [BioC] Package plgem for analysis of spectral counts, was 'limma for spectral counts'

Hi all,
  I am using SAFE for testing gene functional categories for the first time. I have some issues:
(1) my dataset has only small numbers of replicates, 2 to 4. I have therefore used limma for differential expression analysis. SAFE does not include the limma t-statistic, but allows for user-defined (local) statistics. Is it possible to use limma in SAFE, has anyone done this? If not, how bad is it to use the standard t-stat? I've compared limma with standard t (with low variance pre-filtering) and limma clearly had more significant (FDR) results. 
(2) Alternatively, is it possible to just input p-values into SAFE (of course cannot do permutations then)?
Thanks, Ina