Entering edit mode
Hello, I would be happy for clarification regarding shrinkage and gene ranking by P-value. I understand that if using the results function, no shrinkage is being performed. But if using the lfcShrink function either with coef or contrast, the log2FoldChange and lfcSE are affected by shrinkage.
As I understood from the DESeq2 paper (Love et al., 2014) to get the z-statistic, the LFC is divided by its standard error (the stat column in the obtained DataFrame). But, I noticed that the stat and P-values are not affected by shrinkage and the values are identical to the ones I get from results function. Why is it like that? Is there a way I can get the P-values considering the shrinkage? Thank you, Roni
Hello, when I set BetaPrior=T, it also changes some of the genes in my list of top20 DEGs. I suspect this is because setting BetaPrior=T changes the p-values and when I set a FDR threshold , some of my previous top genes get excluded.
Do you think it is appropriate to use both FDR threshold and shrinkage? If yes, which P-values should I use for FDR threshold: the ones before shrinkage or after?
betaPrior=TRUE is deprecated since ~2016. You should use lfcShrink instead (see vignette).