[BioC] edgeR: counts or pseudo.counts?

[Davide Cittaro]

Hello list,
I'm trying to use read counts normalized with edgeR outside R. After a DGEList has been normalized by TMM and dispersion has been estimate (at least with estimateCommonDisp) I have at least three ways to export data:
cpm(y)
cpm(y, normalized.lib.size=T)
y$pseudo.counts

Which would be the most appropriate? Actually I'm having really good results with pseudo.counts, but I would like to know your opinion about.

Thanks

d