simulations and standard errors

[nlsanto2]

Hello everyone,

I am working on MARK simulations with closed models (full likelihood heterogeneity p, pi and c). I am trying to evaluate what is the minimum number of occasions that I need for my study.

To simulate capture histories I used model Mth (2 mixtures) with full interactions and I run 1000 simulations with 4, 6 and 8 capture occasions) also trying increasing values of the capture probabilities.
To estimate N I used the same model as the true model.
My problem is that in some cases (6 occasions-high capture probability and all the cases with 8 occasions) the standard errors of the estimated population size are very small (like 1 or 2).

In this way even if I have pretty accurate point estimates for many simulation runs (e.g. estimated N = 680, true N = 700), the confidence interval does not cover the true population size. I used the 2nd part option for the estimation of variance as recommended in the manual.

I was wondering it is normal to obtain such small standard errors and if there is any alternative way to obtain a more realistic standard error and confidence interval for the N estimates of the simulated data sets.

Thank you in advance,
Nina