relationship of PPEE, PPDE, FDR

[s...@soe.ucsc.edu]

Dear Ning,

Thank you for your efforts on EBSeq and your helpful responses on this forum.

EBSeq_1.8.0

Can you clarify the relationship among:

PPDE  -- the value returned from EBTest() in the DE column of PPMat or PPMatWith0
PPEE  -- the value returned from EBTest() in the EE column of PPMat or PPMatWith0
FDR    -- the argument for GetDEResults()

I think that:

1-PPDE = PPEE = FDR (at least for FDRmethod="hard")

But I could not find conclusive documentation of that in the functions.

I would also be interested to know if you have an opinion about the relationship
between  FDR in EBSeq and the adjusted p-value (BH corrected p-value, I believe)
reported by DESeq. I understand that EBSeq is a Bayesian method, but I am not
quite as clear as to how that controls for multiple hypothesis testing.

Thanks,
Sol Katzman

[Ning]

Hi sol,

You are right, PPDE indicates the estimated fdr. So a list of genes with PPDE> 1-alpha will be comparable to the list of genes with DESeq adj p value < alpha. EBSeq uses one model to estimate the FDR of all genes simultaneously ( instead of multiple independent tests) so no further multiple test adjustment is needed.

Best,

Ning

--
You received this message because you are subscribed to the Google Groups "EBSeq users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ebseq-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.