In families.R, for the delta models using type = "poisson-link", I believe that both link 1 and link 2 should be forced to be "log", whereas the defaults for these functions use a logit link for the first component. I can make the change but wanted to make sure that this wasn't accounted for elsewhere, and that this change is correct.
It seems like this only affects the printed output, but the conventional vs poisson-link delta models are seemingly being computed appropriately. If so, that's an easy fix I can make.
In families.R, for the delta models using type = "poisson-link", I believe that both link 1 and link 2 should be forced to be "log", whereas the defaults for these functions use a logit link for the first component. I can make the change but wanted to make sure that this wasn't accounted for elsewhere, and that this change is correct.