Closed Robust Multstrata-issues with N

[npseudacris]

Greetings:

I have been working on a S(.) Psi (.) p(t) c(t) N(t) model in a Closed Robust Mulitstrata framework. I have 39 total sample periods, 14 primary periods (years) with 2-3 secondary periods (months) per year and three strata. I have over 4000 marked individuals and according to MARK my effective sample size is 11,573. My initial runs of the model without constraints or fixed parameters resulted in “suspect convergence”. Thinking this was not a good thing, I took the advice of various replies from other questions posted to Phidot and tips in chapters 9 and 14. I reran the model with p=c, fixed three transition probabilities that I know could not occur to zero, and used the Mlogit link for the remaining transition probabilities. I got convergence with this model, but fewer parameters were estimable. Although S and Psi are very similar among runs even those with suspect convergence, my Ns vary a great deal and are much lower than I was expecting. I constrained pt as recommended either by making p=c or fixing pt with an estimate from a p=c model run so that pt never equaled 1. Just for comparison I ran a Closed Robust model with the same data. This model didn’t have any convergence problems and the total population size estimates were closer to what I was expecting. Any ideas to what’s going on here or suggestions on how to improve my N estimates in the Closed Robust Multistrata model? Eventually, I want to add covariates and time to S and Psi to this model, but figured I would need parameters from a simpler model first.

As always, thanks!
Nicole