edgeR: Common dispersion and pseudocounts

[flobpf]

Common dispersion will tell me how much variability I should expect, on an average, in a given gene's mean expression value. Is that the variability due to technical effects or biological effects?

Secondly, how does edgeR go from the common dispersion estimate to getting pseudocounts? For example, if the CD is 0.005 and the readcounts of a gene across 2 replicates and 2 treatments are (r1=100,r2=120,x1=500,x2=200), how does edgeR know whether to increase x2 or decrease x1 and to what extent?

Also, at what step are the normFactors used? Are they used in making pseudocounts?

Thanks