Bootstraps and number of estimated parameters

[caspar]

Dear all,

I have a question regarding GOF using bootstrap simulations. I'm running a fair bit number of Burnham LD models in RMark and am interested in obtaining a c-hat value to evaluate model fit. I have a 45 interval EH, with intervals being constrained to be equal across four season dummies, and am interested in obtaining survival estimates for juveniles and adults (2 age classes) in each season.
I have started with a full factorial design for all four parameters and work myself down to a model that can estimate all parameters (so npar.unadjusted =NULL). The largest (most parameterized) model that can still provide an estimate for all parameters is S(s*a)p(s*a)r(a)F(a) where s=season and a is age. When running the exact same model in MARK, I obtain identical results (great relief) but when doing 100 bootstrap simulations I get models with number of estimated parameters often being less than 20. If ignoring this, the model chat divided by the mean c from the bootstraps equals around 1.7 which i can live with. However I remain concerned about the validity of this approach, since the deviance observed in my original model exceeds the highest simulated deviance by 1000 units or so. Subsequently (I know i should have done this before GOF testing) when looking at the model beta estimates, I see se's of 0 and real parameters estimated at boundary (for S:season4), or very low for p (which is reasonable given the population size of the species in study).
Should I conclude that my data is sparse and that some of the parameters under this model are just about estimable? Or should I assume that this is simply due to the logit link issue? I know that most of my resightings and recoveries (67 and 990 respectively out of a pool of 14000 individuals) occur within months after marking and almost never beyond that. Is this sparsness? (I m fairly new in marc recapture analysis techniques and it’s not yet clear to me how a "good" dataset should look like). Or is there some sort of parameter fixing that needs to be done, given individuals of this species live about 5 years, whereas my EH spans over 12 years, and that this may remove the zero se’s from my p parameters?

Any advice on this matter would be highly appreciated,
Thanks in advance,

Caspar