only one site with detections/probability of false absences

[stephanietodd]

Hello (again!),
This is related to the post
viewtopic.php?f=11&t=2299
In particular, where Jim said: 'It takes at least 2 surveys with detections per season in order to estimate occupancy or detection probabilities' however I am running single season models.

Info:
-Study species is a rock wallaby in north-eastern Australia that I believe has severely declined.
-We surveyed 24 sites using sensor cameras for a up to a period of just over 2 months
-study sites where suitable habitat/areas where they have been recorded previously
-The only site it was detected was on an island. There was only two detections in 62 days, but I can make each 'survey event' several days
-The only site covariate that makes sense to include is the dummy variable 'mainland' where a zero indicates island

Question:
-what is the probability, given the detection history, that it was present at other sites on the mainland but missed (i.e. what is the probability of false negatives at every single site?)

-Is it possible to answer this using occupancy Presence? I'm pretty sure there isn't enough data as there is only one site with detections but I wanted to make sure as I'ts a pretty important- to say that they have actually declined/disappeared and it wasn't just that we missed them. As in the above mentioned and other posts I have issues with massive SE values when I run the models (I assume because estimates are very close to zero)

-If not, might it be possible to answer this question another way? I'm thinking along the lines:
detection P (at the site where it was detected) = 2/62=0.032 therefore at another site, with say 53 trap days, the chance of missing the species given that it is present =(1-0.032)^53=0.176. But then what abut SE's? I know this is probably beyond the scope of this forum, but if you could point me in the right direction I would really appreciate it.

Many thanks,
Stephanie

[stephanietodd]

I just reread my post, and I wanted to clarify something- I am aware that the method of calculating detection probability I described is essentially the same as the way Presence calculates it. What I mean is there a way of only modelling detection, not occupancy from my data?

[bacollier]

Info:
-Study species is a rock wallaby in north-eastern Australia that I believe has severely declined.
-We surveyed 24 sites using sensor cameras for a up to a period of just over 2 months
-study sites where suitable habitat/areas where they have been recorded previously
-The only site it was detected was on an island. There was only two detections in 62 days, but I can make each 'survey event' several days
-The only site covariate that makes sense to include is the dummy variable 'mainland' where a zero indicates island

I saw you posted a bunch of times, with limited response so here is the short answer to most of this:
if I understand this correctly, you surveyed 24 sites over some various sampling occasions with camera surveys, and had one site with positive detections over perhaps 2 sampling occasions (which seem arbitrarily defined if you can 'make' each survey event several days)?

If my above is correct, then any attempts at a occupancy analysis are problematic as you do not have enough data to do anything inferential work in a capture-recapture modeling approach. Sorry.

\bret

[stephanietodd]

Hi Bret,
Thanks for your response
Yes you are correct, at least for one of my study species.
I figured as much unfortunately. I am moving on doing some other things and just describing what we found for that species.
Cheers,
Steph