[BioC] limma and p-values

[Giovanni Coppola]

Hi everybody,
my question is again related to p-values...
Is the topTable function using only 'p.adjust' to adjust the p.values?

#R2.0.1, limma 1.8.13

fit<-eBayes(fit)
#After applying eBayes, fit$p.value contains the p-values associated with 
the modified t-statistics

p.corr.none <- p.adjust (fit$p.value, "none")
p.corr.none.ord<-sort(p.corr.none,decreasing=FALSE,na.last=NA)
p.corr.none.ord<-p.corr.none.ord[1:10]
toptable.none <-topTable(fit,number=10,adjust="none",sort.by="P")

# at this point, 'p.corr.none' = 'toptable.none$P.Value' , whereas...

p.corr.fdr <- p.adjust (fit$p.value, "fdr")
p.corr.fdr.ord<-sort(p.corr.fdr,decreasing=FALSE,na.last=NA)
p.corr.fdr.ord<-p.corr.fdr.ord[1:10]
toptable.fdr <-topTable(fit,number=10,adjust="fdr",sort.by="P")

#... 'p.corr.fdr' and 'toptable.fdr$P.Value' are not the same, because 
'toptable.fdr$P.Value' is much worse...why?

Thanks #and sorry for bringing up p-values again :-)
Giovanni