Deviance = 0 in POPAN but not CJS - help?

[lcp729]

I read on one of the other posts that getting a deviance of zero was possible using the robust design model because the saturated model hadn't been computed yet and some constants were left out for faster computation.
Is something along these lines at play in POPAN also because when I run a particular set of data, I get a deviance of zero? This doesn't happen when I run it through CJS, but I'd like to try to get an estimate of N, if possible.
I've read the guide, and I've looked through the posts, but I can't figure out why I'd get this result.
Any help would be greatly appreciated!
Laura

[gwhite]

Laura:
The saturated model is different for the CJS data type from the POPAN data type because of the use of the leading zeros in the POPAN data type. I'm not sure how to compute the saturated model with the POPAN data type -- maybe Carl has some ideas.
Gary

[cschwarz@stat.sfu.ca]

The usual saturated model in CJS and other models that condition upon an animals first capture, is to have a separate probability for each observed history, i.e. if \omega is the capture history (e.g. 010011) then the P(\omega) is simply n_\omega/n_obs where n_\omega is the number of animals with capture history \omega, and n_obs is the total number of animals observed.

However, this doesn't work for models where abundance is estimated because you need to include the animals with history (000000....), i.e. those animals not observed, which also needs to be estimated.

In practise, I would suggest that you compute a deviance based on conditioning on the observed animals. The easiest way is to use the Link-Barker model (which doesn't estimate abundance) but is "equivalent" to the POPAN model. So if you fit a {p_t, phi_t, pent_t} model in POPAN, look at the deviance of the {p_t, phi_t, f_t} model in Link-Barker model. This can be used to estimate a variance-inflation factor if needed.

It wasn't clear from the posts why the deviance of the POPAN model is of interest?

I'll add a brief write up on this to the chapter on JS estimation that is currently in progress (about 80%-90% complete) which will be a chapter in the Gentle Introduction to MARK.

If you want to see the chapter so far, point your browser to
http://www.stat.sfu.ca/~cschwarz/POPAN/POPANexample/
and look at the file
EstimatingAbundance.pdf


Carl Schwarz.