unestimable parameters in closed capture models??

[jSandrini]

Hello everybody,

I constructed a full closed capture model with heterogenety for a non-invasive genetic mark-recapture dataset for a roe deer population with "sex" as an attribute group.
In this model, the male subdataset had a constant p and c, and the female subdataset had p varying over time.

The problem is that in this model parameter 1: pi (for male subsample) has a standard error of 0.00000
and Parameter 2: pi (for female subsample) has a standard error of 0.00000 too. Correspondingly, both have no real confidence intervals (see below).

Here my question:
Is the model not valid or is it possible that both pi parameters (1&2) aren`t needed for modelling, because the model does not incorporate individual heterogeneity, so that parameter pi does not need to be estimated in that case?
The model should have 10 parameters altogether, but in the AIC count, only 8 are given. Do I have to adjust the parameters in this case, or is it okay because the parameter pi does not have to be estimated?

Here is the model (Male_M0_female_Mt) output of real parameters:


Real Function Parameters of {Male_M0_female_Mt}
95% Confidence Interval
Parameter Estimate Standard Error Lower Upper
------------------------- -------------- -------------- -------------- --------------
1:pi 0.4500732 0.0000000 0.4500732 0.4500732
2:pi 0.4500968 0.0000000 0.4500968 0.4500968
3:p 0.2282437 0.0287038 0.1768784 0.2892830
4:p 0.3247443 0.0461856 0.2414280 0.4208608
5:p 0.2869833 0.0437938 0.2092318 0.3797524
6:p 0.2794311 0.0432876 0.2028362 0.3714715
7:p 0.2492223 0.0411588 0.1774228 0.3381337
8:p 0.1359395 0.0311556 0.0855430 0.2092331
9:N 97.264473 9.1566562 84.523055 122.01085
10:N 132.41187 8.9054664 119.45955 155.55662

Thank you for your time.

Julian

[dhewitt]

Julian,

I have no experience with closed captures models with heterogeneity, but the pi estimates you got are just the default starting values in MARK's optimization routine. In other words, MARK did nothing with those parameters in the optimization. My guess is that the model is not set up properly, but I suppose the result could also mean that there is no information about heterogeneity in the data set. Hopefully others with experience with heterogeneity mixture models will chime in.

- Dave