additive time effects in recoveries in Gemaco

[Ginger]

I have a combination of live resightings and dead recoveries. I have 4 groups and defined 2 states (live and dead).
Individuals in all groups can be resighted alive, but only members of groups 1 & 2 can be recovered dead.
There is only 1 age class. I'm using a CAS model with separate survival/transition parameters in M-Surge.
I coded survival as 'f' and interpret transition as survival (I fixed the transition from(dead)to(live) to 0). I changed the transitions pattern matrix to
|1 0|
|1 0|

I want to run a model with additive effects of group and time (no interaction).
I coded p in Gemaco as

[g+t].to(1)+[g(1,2)+t].to(2)

I get separate estimates of p for live resightings for the 4 groups that vary in parallel over time, as expected.
I expected to only get estimates of p for dead recoveries for groups 1 & 2. However, I also get estimates for group 3 in the reduced set of parameters. Occasions 1-11 are 0, and I fixed the last occasion (13) in both p and psi, but occasion 12 is non-zero.

How should I code p to get dead recovery probabilities just for groups 1 & 2?

Thanks and merci beaucoup in advance for any advice.

Ginger

[CHOQUET]

Hello,

I don't understand the structure of your transition matrix.
Rather consider the model described in the document

http://www.cefe.cnrs.fr/BIOM/pdf/Choque ... models.pdf

page 9.

The only difference between this example and your study is that
in your case, the capture rate is not zero.

to get the appropriate capture rate, use the following sentence in GEMACO:

[g+t].to(1)+g(1 2).[g+t].to(2)+g(3 4).to(2)

The second term "g(1 2).[g+t].to(2)" is an additive model between group and time that applies only to g(1.2).to(2).

The last mathematical parameter link to "g(3 4).to(2)" has to be fixed to zero because for group 3 and 4, the recovery rate is zero.

Sincerely,
Rémi

[Ginger]

Remi,

Thanks so much for your detailed reply. I'm excited to get on with this analysis.

I don't really understand the transition matrix structure either. I attended the Analysis of Multistate Capture-Recapture Data workshop in Quebec in 2006, and I have a note in my answers to exercises book to change the matrix for dead recoveries analyses. Perhaps I misunderstood that. I haven't had a chance to work with M-Surge since then, so I don't remember some of the details very well.

Merci beaucoup,

Ginger