DESeq power calculation

[ddf555]

We're trying to calculate the effect size cutoff for a given power — in this case we have 4 naïve samples, 4 experiment samples, and we're using DESeq for the differential expression analysis.

So running these parameters through a power calculator -- we're leaving delta empty, so the system can tell us what the effect size should be.

pwrCalc<-power.t.test(n=4,sig.level=0.01,power=0.9,sd=???,delta=NULL)

So, the question is how do we get the sd out of DESeq — with that, we can calculate the delta that is required for a power 0.9 and sig level of 0.01.

Please let us know if anyone has a suggestion, or if there's a different way to calculate the power, or if we're thinking about this wrong. Thanks!

[dpryan]

The variance is mean dependent and, anyway, this would only give you power for one gene. Further, you appear to be calculating post-hoc power, which is typically meaningless. The question you have to ask is what you're trying to achieve. Typically, you want to know how many samples you would need to have a reasonable power of finding reasonable magnitude fold-changes, for which tools such as Scotty would seem relevant.