Potential New Feature: log-base-2 link function

[ajinkya-k]

Differential gene expression models use log with a base of 2 as the link function to model the mean in negative binomial models. For instance see the DESeq2 vignette theory section.

I propose adding Log2Link. (Additionally we could also add size factors as an option, though im not sure if the same thing can be handled with weights argument in the negbin instead)
Does that seem like a good idea?

[andreasnoack]

Wouldn't it just be a matter of scaling the parameters by $log(2)$? If so then I'm not sure it is worthwhile to add the extra definitions.