MSORD and unobservable states into secondary periods

[Javier Klaich]

Hi everybody!

I have been collecting photo-id data from a local population dolphin. The seasonal sampling follows the typical Pollock`s robust design. The seasons represent the primary periods and k days for each season represent the secondary periods. These secondary periods are no longer than 20 days for each. My study area is not enough to cover the real biological area, therefore in secondary periods there is at least one unobservable state (no geographic closure of course). I decided to use MSORD to analyze my data in order to model unobservable state for each secondary period and to study the temporary emigrations between seasons by means of gammas estimation. This (I thought) implies to possibility to estimate Psi parameters for every single secondary periods (full time or not effects). The objective is to take into account the movement in and out side the study area in secondary periods. The problem is when a look the model parameters structure, I see just the possibility to estimate Psi between Primary periods and not into the secondary periods. The pent and Phi parameters do not solve the problem because once the animal did enter to the study area, the possibility to come out and return is not allow in the MSORD. My first idea was to create an encounter history like:
UU1U 000 1U11
00000 1UU 0000
When the animal is detected at least once in a secondary period the events of no capture are coded with U (for unobservable but in the super-population) but if the animal is not detected at all into a secondary period is coded with 0. The objective of this was to weight the fact that the animal never was detected and is more likely an temporary emigrant than a non-detected with probability 1-pi
Please, if anybody has a clue for me I will appreciate it.