Pooling Parameters CJS/POPAN

[Fish_Boy]

Is there any literature on the impacts of or acceptance of pooling parameters that are essentially the same? For example in the outputs below there are three sample periods (3 Phi(.) estimates) and a series of recaptures for all the remaining p(t). My question is parameters 7 and 12 are the same, so is it a faux pas to make both parameter 7 in the parameter index matrix?

Real Function Parameters of {TEST_2}
95% Confidence Interval
Parameter Estimate Standard Error Lower Upper
-------------------------- -------------- -------------- -------------- --------------
1:Phi 0.9538172 0.0338082 0.8210634 0.9893571
2:Phi 0.7435387 0.0416695 0.6538809 0.8164897
3:Phi 0.9179799 0.0136872 0.8868447 0.9411169
4:p 0.0428398 0.0308482 0.0101404 0.1635602
5:p 0.2052694 0.0414394 0.1356898 0.2982177
6:p 0.2800241 0.0462829 0.1987237 0.3788591
7:p 0.1911235 0.0364782 0.1295203 0.2728430
8:p 0.2602866 0.0402284 0.1893870 0.3463866
9:p 0.1113749 0.0210120 0.0763579 0.1596743
10:p 0.0826173 0.0189370 0.0523009 0.1281309
11:p 0.0730903 0.0180083 0.0447383 0.1172051
12:p 0.1916858 0.0302490 0.1392253 0.2579889
13:p 0.2368790 0.0327524 0.1787278 0.3068805
14:p 0.1504001 0.0238804 0.1093143 0.2034016

[ehileman]

In this particular case, I don't see any problem with what you are suggesting. But why bother? My guess is it will only trivially improve the precision of your estimates, which are already decent, and require you to provide some kind of a justification in your methods section for the post-hoc analysis. Hardly seems worth it to me. My two cents.

[simone77]

Not sure it is a good idea. Pooling parameters make sense to me if you have some a priori expectation about them, and it represents a different hypothesis. To me, a capture-recapture model is the quantitative formulation of a hypothesis and, as such, produces estimates (no vice-versa). I do not understand well your system study because I don’t see how you have 3 capture sessions, 3 phi estimates and 11 p (recapture) estimates. However, plotting the p estimates I do not see any pattern that helps me to understand why p 7 and p 12 should be, a priori, the same.
Also, I agree with ehileman that, from a practical point of view, it does not seem to worth it either.

[Fish_Boy]

Thanks for the responses.

I do not understand well your system study because I don’t see how you have 3 capture sessions, 3 phi estimates and 11 p (recapture) estimates...

The 3 Phi estimates are for three different periods over 24 years... Example the hypothesis being that survival was impacted by different levels of exploitation e.g., Phi2. a reduction in harvest pressure since 2006 has produced an increased survival.

And thanks for the perspective

To me, a capture-recapture model is the quantitative formulation of a hypothesis and, as such, produces estimates (no vice-versa).

Cheers.