I'm using the VGAM package in R with the gamma2
family function (two parameters: mu and shape).
I'm hesitating in the choice of the link function. In fact, I have the two following fits:
fit1 <- vglm(y ~ x,
family = gamma2(lmu ="identitylink", lshape="identitylink"),
trace = TRUE,
data = gdata,
na.action = "na.omit",
maxit = 100)
fit2 <- vglm(y ~ x,
family = gamma2(lmu ="loge", lshape="loge"),
trace = TRUE,
data = gdata,
na.action = "na.omit",
maxit = 100)
The two fits had no problem converge, even if fit2
converges faster than fit1. Which one to choose, does it makes sense to compare AIC(fit1)
and AIC(fit2)
- of course on the exact same dataset - or do I have to rely on other criteria?