Recaptures only VS. POPAN

[Luigidigidoo]

Hello everybody!
I have a file about capture-recapture of birds during the pre-nest season. The survey tooks 8 years, one day per year, and only live animals were taken on account. We analyzed it with the "recaptures only" and the "POPAN" models, and we are using the N from the POPAN and the phi and p from the recaptures only.
I think that phi and p from POPAN should be useful too, but no one is agree with it.
The question is, there is some reason for not to use the phi and p from POPAN? POPAN don't calculate phi & p like a CJS model? Where is the explanation of the "recaptures only" model in the MARK book?
Thanks in advance

[jlaake]

The term "recaptures only" refers to a CJS (Cormack-Jolly-Seber) model which only models the recapture events. The initial "1" and any leading 0's are not included in the likelihood -- thus "recaptures only".

POPAN is one particular formulation of a JS (Jolly-Seber) model. JS models to include the initial 1 (first capture) and any leading 0's.

If you are fitting the POPAN model to get N, you should use the Phi and p estimates from that model but be aware that some parameters are confounded and not estimable depending on the model you specified.

In the documentation, anywhere you see CJS it is recaptures only. There is also a chapter on POPAN and the other Jolly-Seber models.

[Luigidigidoo]

Thank you, jlaake, but with your explanation, do you mean that the initial p & phi are also confounded and not useful in the recaptures only?

[Luigidigidoo]

By the way, if phi and p are constant in POPAN, but pent are time-dependent ¿Does it means that final survival and catchability are not estimable?
Thanks again

[jlaake]

You should read the sections on CJS and JS before you dive into this. Parameter confounding depends on the specified model. For CJS the classic example is with the model Phi(t)p(t), where the both Phi and p for the last interval/occasion are not estimable. However, with Phi(t)p(.) they are all estimable. I don't know of any cases where Phi and p are not estimable for the first interval/occasion for CJS. Recognize that CJS does not estimate p for pccasion 1 whereas that parameter is included in the JS (POPAN) model. The chapter in the book on Jolly-Seber models explains which parameters are estimable and when.

--jeff

[cooch]

You should read the sections on CJS and JS before you dive into this...

At least. Given the level of your questions, you are strongly urged to read (study) Chapters 1 -> 7, all of which have to do with CJS models. Reading them will help you understand in general which parameters are confounded in the CJS context, why, and what you can do about it. Then, after reading those chapters, and depending on your purposes, you might look at the chapter on JS models (which ostensibly is the 'POPAN chapter').