what is the function of prior.count in lcpm ?

Question

[Björn]

In this command line while using edgeR, lcpm<-cpm(y,prior.count=5, log=T) what is the applicability of prior.count there ? I found "prior.count as average count to be added to each observation to avoid taking log of zero. Used only if log=TRUE." Is there any range to be used ? How to know which threshold is right ?

[Zhilong Jia]

Log zero will -Inf. Adding a prior.count will make the logged value be countable. The prior.count should be small, like the default 0.25 or the min of y except 0 to weaken the effects on y.