Every glm() family already carries its link and inverse-link
functions (stats::family(mod)$linkfun / stats::family(mod)$linkinv),
but many users don't realise these are available for the taking.
erglm_link() and erglm_invlink() are thin, discoverable wrappers
around them: erglm_link() maps the response scale to the linear
predictor scale, and erglm_invlink() maps the linear predictor
scale back to the response scale.
Examples
mod <- erglm_model(ae1 ~ aucss + sex, erglm_data, family = binomial())
erglm_link(mod)(0.5)
#> [1] 0
erglm_invlink(mod)(0)
#> [1] 0.5
erglm_link(mod)(erglm_invlink(mod)(-2:2))
#> [1] -2 -1 0 1 2