Wouldn't it make sense to document the aic function in family
I'm currently writing an R-program which makes use of the
log-likelihood function of e.g. a logistic regression model or a
poisson regression model. As I have observed, the aic function
in a family object calculates the part of the AIC which does not
depend on the predictor dimension.
As this is not documented I should probably not use it. Or is
there another solution without rewriting the already implemented