GOF for Burnham's live and dead encounters

[hhh5]

Hi
I am trying to estimate immature survival in Atlantic puffins, using Burnham's both live and dead encounters. I use dataset containing birds ringed as breeders which are mostly recaptured breeding and individuals ringed as nestlings which are mostly recovered through harvest.

What would be the most suitable way to test GOF?

I have tried modeling the breeders using CJS model and tested GOF using U-CARE, which reviled a strong transience effect, so I used Pradel's et al (1997) method Phi(a2 t/t)p(t) . I tried incorporating this into Burnham's approach by using the model S(t)p(t)r(t)F(a2 ./.) and it resulted in a better fit and LRT shows:

Reduced Model General Model Chi-sq. df Prob.
------------------------- ------------------------- ---------- --- ------
{S(t) p(t) r(t) F(.) PIM} {S(t) p(t) r(t) F(a2 ./.) 103,241 6 <.0001


I have not however found anything in the literature allowing me to do this and statistics is regrettably not my strong side. My reasoning for doing this is that Phi=S*F and transience affects site fidelity not survival.


I found the preliminary results quite intriguing and if anyone is interested:

F transients = 0.6592089 SE = 0.0231684
F residents = 0.9672762 SE = 0.0096561

while for the corresponding CJS model Phi(a2 ./.)p(t)

Phi transients = 0.6488289 SE = 0.0211095
Phi residents = 0.8824343 SE = 0.0034656

[hhh5]

I should maybe also note that the harvest i.e. recoveries and recaptures coincide both in time and space, i.e. during the summer in the colonies.

[egc]

I should maybe also note that the harvest i.e. recoveries and recaptures coincide both in time and space, i.e. during the summer in the colonies.

If dead recoveries only occur where live encounters occur. From Chapter 9:

If all dead recoveries and live encounters occur in the same sampling area, then clearly realized fidelity F for the marked
sample is 1 - so, you can use the same approach, except you fix F=1. But, in general, you need to think
hard about what the parameters mean, and how the sampling assumptions (and whether or not they
are met) influence the interpretation of the parameter estimates.

[hhh5]

Hi Thanks for your reply

I had seen this and I take this into account for the immatures, F fixed as 1, but since there are very few recoveries of breeders I did not think it would matter, however I know there is an transience effect there and I wonder if should or can, try to take that into account e.g. by the aforementioned model?

[Morten Frederiksen]

Dear hhh5,

Your approach seems sensible. As far as I understand the study, the marking and live recaptures only cover a small part of a huge colony, whereas the dead recoveries (harvest) cover a larger part of the colony. In this situation, you would expect transience/permanent emigration in the data, even though established breeders probably rarely (if ever) move to another colony. This within-colony emigration should be well captured by F in the live-dead encounter framework. Your estimate of F for the second age class is close to 1, indicating that movements mainly occur during the first year after marking - perhaps because newly marked birds are mainly recent recruits which might be more prone to move. You could try to fix F to 1 for the second age class and see how much deviance increases. If there is permanent emigration out of the colony, you will of course not be able to estimate it due to the lack of dead recoveries elsewhere.

Regarding goodness of fit, you can use the median c-hat approach in MARK for this data type. It will give you an overall assessment of fit, but without the detail available from U-CARE.

MOrten

[hhh5]

Thank you very much Morten.

You assume correctly, the recaptures happen only in a small part of the colony and further more only in burrows shallow enough for the ringer to reach into, so to cause transience an individual would only need to move few meters into a slightly deeper burrow or even dig a little bit deeper. I think since birds cannot be be re-sighted but must be recaptured to be read and puffins are apparently very sensitive to disturbance it is not far fetched to think of this as a possible handling effect.

Vennlig hilsen
Dáni