psi estimates: missing data and all sites occupied

[tpinn]

We are running a single-season model with spatial correlation, and we have 50 surveys (spatial replicates) for most of our 24 sites. We have two questions:
1. Presence is not giving us an estimate of psi for one of the sites in which we are missing data for the first survey. We ARE getting estimates of psi for other sites with missing data, as long as the data are not missing for the first survey. Does anyone know what we can do to get an estimate of psi for the site that is missing data in the first survey?
2. For our 24 sites, the species occurred on at least one survey for each site, which means that the species was present at every site. PRESENCE has estimated psi to be 1.0 for all of the sites. To get more realistic estimates of psi for sites (i.e. something other than 1.0), should we include sites in the analysis where the species was never detected at any of the 50 surveys for a site?

[jhines]

I thought I responded to this earlier, but I don't see my response here, so I'll post now...

I'll answer #2 first. If you have detections at every site, then the estimate of psi should be 1.0. That would be the best, or most realistic estimate, given the data. If you've withheld data where some sites had no detections, then you're not estimating occupancy for all sites, just the ones you included. So, the answer is yes, all sites should be included, not just the ones with detections. The way the model works is that detection probability can be estimated from the sites with at least 1 detection because you know that 1's in the detection history represent detections (p), and zeros represent non-detections (1-p) of occupied sites. Once p is determined, they can be applied to the other sites (with no detections) in order to figure out what proportion of those are really unoccupied, as opposed to occupied-but-not-detected-in-each-survey.

#1. I don't think I've tried analyzing this type of data with a missing value at the start, but it should give estimates for all sites (if you have site covariates), or a single estimate if no covariates. I've updated this model recently, so my suggestion is to try your data again with the most recent version of PRESENCE and if you still don't get an estimate for that site, send me the most recent backup zipfile from your project folder.

Jim