To pool observable/unobservable states in E-SURGE

[pamela]

Hello,

I would like to pool states to consider them equal.

Description: I work with 6 states (1- breeding in A, 2- breeding in B, 3- breeding failure in A, 4- breeding failure in B, 5- not breeding in A or B, 6- and death) and 5 events (incubation in A, incubation in B, rearing in A, rearing in B, and NO).
I have 3 capture sessions per year: the first with event incubation, the second with incubation and rearing, and the third only with rearing. States “failure” and “not breeding” are unobservable. To consider a failure (in session 2 or 3), there is automatically event “incubation” or “rearing” before (in session 1 or 2).

I want test hypothesis “dispersal depends on site, no breeding success”. I pool states 1-3 for site A and states 2-4 for site B. In GEMACO, I write f(1 3,2 4,5).to(1,2,5). Normally, in excel results file, I can have estimates: 1 to1 equal to 3 to 1, 1 to 2 equal to 3 to 2, 1 to 5 equal to 3 to 5, 2 to 1 equal to 4 to 1, 2 to 2 equal to 4 to 2, 2 to 5 equal to 4 to 5. It is not the case, I have an estimate for each state, as if I wrote f(1_5).to (1,2,5).
Secondly I want test hypothesis “dispersal depends on breeding success, no site”. I pool states 1-2 for breeding (observable) and states 3-4 for failure (unobservable). In GEMACO, I write f(1 2,3 4,5).to(1,2,5). This time, I have good estimates (equal) for states 3 and 4 to 1 or 2 or 5, but bad estimates for states 1 and 2 to 1 or 2 or 5.
Why E-SURGE can pool states 3 with 4, but not the others? We can pool only unobservable states or stat? Perhaps I don’t speak very well GEMACO language…

Thanks in advance,

Paméla

[CHOQUET]

Hello Pamela,

Concerning

>Description: I work with 6 states (1- breeding in A, 2- breeding in B, 3- breeding failure in A, 4- >breeding failure in B, 5- not breeding in A or B, 6- and death) and 5 events (incubation in A, incubation >in B, rearing in A, rearing in B, and NO).
>I have 3 capture sessions per year: the first with event incubation, the second with incubation and >rearing, and the third only with rearing. States “failure” and “not breeding” are unobservable. To <consider a failure (in session 2 or 3), there is automatically event “incubation” or “rearing” before (in >session 1 or 2).

>I want test hypothesis “dispersal depends on site, no breeding success”. I pool states 1-3 for site A and
>states 2-4 for site B. In GEMACO, I write f(1 3,2 4,5).to(1,2,5). Normally, in excel results file, I can
>have estimates: 1 to1 equal to 3 to 1, 1 to 2 equal to 3 to 2, 1 to 5 equal to 3 to 5, 2 to 1 equal to 4 to
>1, 2 to 2 equal to 4 to 2, 2 to 5 equal to 4 to 5. It is not the case, I have an estimate for each state, as
>if I wrote f(1_5).to (1,2,5).

Constraint of equality between row is preserved when the number of parameter to estimate is 1 on each row, this is not the case here.
This is not a problem of syntax in Gemaco but rather the fact that the multinomial logit is used.
One way to deal with this problem is adressed in:
Grosbois, V. and G. Tavecchia (2003). "Modeling dispersal with capture-recapture data: Disentangling decisions of leaving and settlement." Ecology 84(5): 1225-1236.
and its implementation in
Choquet, R., L. Rouan, et al. (2009). Program {E-SURGE}: A software application for fitting multievent models. Modeling Demographic Processes in Marked Populations. D. L. Thomson, E. G. Cooch and M. J. Conroy. Dunedin, Springer. 3: 845-865.

A similar problem was described in the forum.

>Secondly I want test hypothesis “dispersal depends on breeding success, no site”. I pool states 1-2 for >breeding (observable) and states 3-4 for failure (unobservable). In GEMACO, I write f(1 2,3 4,5).to
>(1,2,5). This time, I have good estimates (equal) for states 3 and 4 to 1 or 2 or 5, but bad estimates for >states 1 and 2 to 1 or 2 or 5.
>Why E-SURGE can pool states 3 with 4, but not the others? We can pool only unobservable states or >stat? Perhaps I don’t speak very well GEMACO language…

This is the same problem than previously.

Sincerely,
Rémi Choquet

[pamela]

Thanks,
I think I have solved the question. Articles are very interesting and have clarified my problem.
Finally, I divided in 3 breeding states: one breeding state by session, because transitions are not the same between sessions (consequently time dependant). Constraints imposed previously for each session don’t allow a sum of 1 for each row in matrix movement.
That’s good until the next error

Paméla