search

fhui28 / CBFM

Spatio-temporal joint species distribution modeling using community-level basis functions
13 stars 2 forks source link

Zero inflated distributions #3

Closed fhui28 closed 3 years ago

fhui28 commented 3 years ago

ZIP/ZINB with just a specifc-specific zero inflation probability? Hard to do with smooths? Well with constant species-specific probabilities of zero it may not be too bad...see https://github.com/AustralianAntarcticDivision/zigam

chrishaak commented 3 years ago

For ZINB and/or hurdle models implemented in a GAM framework, maybe also see the zinegbinomial and zanegbinomial functions in the VGAM package?

fhui28 commented 3 years ago

Thanks Chris,

This is an option, although I am trying to reduce having to make a dependence on VGAM just for this, and given I am using mgcv for all other response distributions currently.

I do not think it is that hard to implement in principle, but I will need to spend some time coding this up...probably start with ZIP first and then move to ZINB...

fhui28 commented 3 years ago

ZIP added

fhui28 commented 3 years ago

ZINB should be added