search

tslumley / svyvgam

Design-based inference for VGAM
6 stars 2 forks source link

AIC() and anova() for svy_vgam models #1

Open coreysparks opened 2 years ago

coreysparks commented 2 years ago

Thanks for porting the VGAM models to use your survey package. Any plans on allowing survey::AIC() or survey::anova() to accept these svy_vglm() models?

tslumley commented 2 years ago

I'll put it on the list. AIC will be a bit more tricky than anova, because it's necessary to worry about the influence functions for nuisance parameters (like the dispersion parameter in a Normal linear model), but that shouldn't be impossible. Anova for symbolically nested models should be pretty straightforward (and thus regTermTest as well)

symbalex commented 2 years ago

Seconded on AIC, if you get the chance! Thank you :-)