from what I understand then (note: this is a follow up to a question asked on the tximport github) the cpm counts calculated in the last line of the tximport edgeR vignette, that is:
cpms <- edgeR::cpm(y, offset = y$offset, log = FALSE)
cannot be used directly (I mean as they are) to calculate logFC between expression levels of a gene of interest in two different samples (say I have only two samples, pre and post-treatment). Instead one should further process y using something like:
y <- estimateDisp(y) et <- exactTest(y)
In other words, neither the cmps calculated as above nor the count values found in y can be used as they are for between samples comparisons, is that correct?
Thank you very much for your help with this matter! Marco