Huggins and Nonidentifiable p

[wsbeatty]

Hello All -

I have searched for this scenario in numerous sources and cannot find a definitive answer. I have run the Mtb model with Huggins Closed Capture [p(i), c(i)] with the appropriate constraints. However, one p(i) is nonidentifiable. [Specifically, in my dataset it is p(7) out of a total of 10 sampling occasions.]

So, given the Huggins model, is the derived N still unbiased if there is at least one p(i) that is not identifiable? I have noticed that confidence intervals for N in models with at least one nonidentifiable p are relatively large compared to a model with p(.) where there are no identifiability issues. Does the larger confidence interval sort of account for the non-identifiable p(7)?

Thank you,

Bill