Using psi sub c as a covariate input in a different model

[amdurso]

My project involves using the presence or absence of different prey species as a site covariate for estimating occupancy and detection of my species of interest (aquatic snakes). Detectability of these prey groups (fish, crayfish, etc...) is quite high (96-100%), but I feel it's contrary to the whole exercise to use catch per unit effort or something similar when it's possible I could have missed detecting some prey groups at some sites.

It was recommended to me that I use Ψc from chapter 4 of the MacKenzie book, Ψc = Ψ(1-p)^K / (1-Ψ)(1-(1-p)^K), which is working really well. My problem comes when I want to incorporate the estimate of error into the error estimated by the larger model into which I'm putting Ψc as a covar. I was told the variance-covariance matrix in the PRESENCE output file is the place to look, but I can't find any instruction on what to do. Any suggestions?

[darryl]

Hi There,
I suggest you look into the two-species models (Chap as there you're explicitly accounting for the imperfect detection of both species.

If your detection of the other species is really that high, then I imagine your estimates of psi_sub_c should be essentially 0 or 1 and have really small SE's (almost 0), in which case it may not be too much of a crime to not worry about trying to include the estimate of the error in psi_sub_c.

Darryl

[amdurso]

Ok, thank you. The psi_sub_c estimates are indeed very near one or zero - I avoided using 1 and 0 because they created many singularities in my data and caused some of my model parameters to be inestimable. The errors on the psi_sub_c's are also quite small.