search

hmsc-r / HMSC

GNU General Public License v3.0
102 stars 37 forks source link

Testing for Zero Inflation for the Hurdle Model #76

Open aeiche01 opened 3 years ago

aeiche01 commented 3 years ago

Super quick question on testing for zero inflation to see whether to use a lognormal poisson or hurdle model for count data. If we have multiple species, do we test for zero inflation on all the species counts lumped together or do we test for zero inflation individually for each species? And if we test for zero inflation individually, do we have to move to the hurdle model if only a small number of species are zero inflated?

ovaskain commented 3 years ago

Hi,

This is not an easy question / there is no straightforward "do this" solution. At least in theory the best would be test individually, and then apply the hurdle/log-normal Poisson model only to those species for which it is needed. In practice probably easiest to take the same model for all species.

O2

On 14.12.2020 22.08, aeiche01 wrote:

Super quick question on testing for zero inflation to see whether to use a lognormal poisson or hurdle model for count data. If we have multiple species, do we test for zero inflation on all the species counts lumped together or do we test for zero inflation individually for each species? And if we test for zero inflation individually, do we have to move to the hurdle model if only a small number of species are zero inflated?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hmsc-r/HMSC/issues/76, or unsubscribe https://github.com/notifications/unsubscribe-auth/AEIYMZTCIKUNJ2SALTSGLE3SUZWCXANCNFSM4U3IQHQA.