My question may be of technical nature: I am trying to model disease counts (d) by using population (p) as offset to control for exposure. In R, I found two possible ways to go:
m1 -> glm(d ~ 1 + offset(log(n)), family=poisson, data=dat)
m2 -> glm(d ~ 1, family=poisson, data=dat, offset=log(n))
The summary of m1 and m2 shows that summary(m1)
= summary(m2)
but if I try to calculate the McFadden through the pR2
(pscl package): McFadden(m1)
≠ McFadden(m2)
.
Does someone have an explanation for that?
Here is the source code of
pscl:::pR2.glm
:If the offset is specified in the formula, it gets lost in the second line (
update
to compute the intercept-only model).See this example: