Barker Combined Data model

[conroy]

I'm using the Barker combined recapture-resighting-recovery model to estimate survival rates of a bird that seems to have fairly high rates of dispersal-- so, 6 month apparent survival rates are fairly low. The data structure actually has few (essentially zero ) recoveries, but there are substantial numbers of resighted birds over a fairly large area (resighting is via metal tags visible through scopes or binoculars). The Barker model seems to work, and provided estimates of S that are reasonable, after contstaining the recovery rates to be zero.

My question is: Is it reasonable to interpret S as 'true' survival, if there are no recoveries in the data (again, the model is constrained as above)? Likewise, what is a reasonable interpretation of the F and F' parameters? I will not claim that sampling for resighting encompasses the entire range that birds can travel inter-periods, but there are a finite number of large colonies (this is an urban, invasive species) that are readily identifiable and countable, so I think it's fairly complete sampling.

Thanks

Mike Conroy

[Bill Kendall]

Mike,

Richard can chime in, but my understanding is that an incidental observation is treated as a recovery that does not die. To rigorously separate S and F requires universal sampling of the animal's range by the incidental observations. So it sounds like you've almost achieved that, with some uncertainty. Therefore, as you suspected, S might still be slightly negatively biased. Since you describe discrete colonies, I assume you are not using MS models because sampling in those other colonies does not fall within a well-defined sampling interval?