Dear Ken,
I am using the ubms package to estimate the occupancy pattern of rare species (which means low detection data), and the package is working very well. I am super happy with it.
When I compare the candidate models by using modSel(), a concern comes to me. I searched around but failed, so I hope you can help to solve it.
As you might know, in unmarked package, we have the ΔAIC for each candidate model. Models within ΔAIC of 2 (or say with ΔAIC<2) are considered to have the same support as the top model (which has lowest AIC). By this way, I take all models which have ΔAIC<2 into further running.
In the meanwhile, the ubms package ranks the candidate models by using elpd. And you also pointed out that the comparison of elpf_diff and se_diff can help to indicate the models with higher elpd are performed better. The question came to me at this step: which models I should select for further running? All of my models have elpf_diff higher than se_diff at least 2 or 3 times. And I could not take them all in the next running.
So my question is: do we have any kind of that ΔAIC<2 "threshold" for ubms package?
Please let me know if my phrase was not clear, or if you need more information from me.
Thank you and best regards,
Thanh