POPAN Models and Adjustment of c-hat

[ADoug]

Hi everybody,

I have four datasets that I have run as CJS data type, building models with constant or temporal variation in p and phi parameters. Then I ran the GOF – RELEASE test within MARK, which gave me significant results for test 3.SR for all datasets, but no significant results for the other tests. Looking carefully to the contingency tables for each occasion I found out that the occasions with excessive number of transients ( 2 or 3 occasions out of 10) were the ones with significant results for the test 3.SR. The next step was include TSM models. As expected TSM are the models with lowest AIC, specifically model p(.) phi (M2 - ./.) for some of the datasets or p(t) phi (M2 - ./.) for others.

But my interest in the study is population size, so I had to run POPAN.
POPAN data type is not separately in cohorts, so it is not possible to built TSM models. My results had lowest AIC for models p(t) phi(.) or p(.) phi(t) depending on data set. The only option that I have for a calculation of c-hat is the GOF – RELEASE (the very same used in CJS, which I know is performed only for the saturated models).
The c-hat calculated using GOF-RELEASE (chi-square value /degrees of freedom) is above the value of 1 (but no bigger than 2). My question: Is it proper, in that case, to adjust the c-hat value calculated before pick up the best model that fits the data?

If I adjust it, models with constant survival is best ranked and it seems more realistic then models with temporal variation in survival since the study was done within a year and the species is a long lived mammal, so this is a very short period to consider significant trends in deaths. Besides permanent emigration, also temporary emigration was very likely to happen during the study and maybe some tendency of animals associated more with specific individuals then others. So I do believe that some lack of fit occurred due to extra-binominal noise, despite the structural problem of the models in POPAN. I am not confident in adjust c-hat because as I read in chapter 5 of “Gentle Introduction” if there are structural problems in the model adjust c-hat might be inappropriate.

Can someone help me to solve this issue?
Thank you,
Alex