Estimatable parameters

[annaren]

1)When I run the closed population model with all my sites because of low population sizes in some areas MARK can not estimate all the parameters so for example in one model 30 out of the 32 parameters were estimatable. The two parameters that could not be estimated were N for an area with 4 and one with 10 - how ever it could estimate one for N=8.
So I wondered if by having unestimatable parameters affected how MARK estimated the other parameters so I took out the two offending sites and ran the model. This time the model estimated all parameters (including the site with 8 individuals). I looked at the estimates that it gave for each parameter and they are very very similar. As the model including the unestimatable parameters were exactly Mt+1 (the number trapped) I was wondering what the best 'statistical' way to procede:
a) Include all sites and present the model with the lowest AIC and most parameters in model estimated as possible?
b) Take out offending sites as they don’t gain anything from the model but may slightly affect the power given to estimate the parameters from the other sites?

Thanks
Anna