www.phidot.org

I have mark-resight data on the life circle of a shorebird species. I am interested in testing differences in survival of birds migrating through geographically distinct staging areas and transition probabilities between them.
The challenging part is the structure of data:
-For each year there are 4 more or less distinct primary sessions (of uneven length) in different geographical (3-4 stratas at each time slice) points.
- We may assume that the population is closed (though in some of the stratas the resighting probabilities is very low).
As far as multistrata-mark-resight models do not seem to be available. What are the options you would recommend to use?
I can imagine:
1 – Use of regular multistrata
2 – Multistrata RD?
3 – Work on the implementation of mark-resight multistrata? Is there any interest from gurus?
Any ideas are welcome!
Eldar

Hi Eldar,

It appears the gurus are busy, so I guess you're stuck with me. If you're only interested in strata-specific survival and transition probabilities, then just apply the regular multistrata or multstrata RD using your marked individual encounter data. If you were interested in estimating abundance as well, then you might want to consider using a multistrata mark-resight model that includes your unmarked individual sightings. To my knowledge, you currently can't do this in MARK (unless someone, e.g., Bill Kendall, has figured out a way to "trick" MARK's multistrata RD models to include a simple abundance estimator; see Kendall et al. 2009, Ecol. Appl. 19: 55-68).

Cheers,
Brett

Thanks Brett!
I have two more questions:
1. With any model I will go I have an issue of defining length of unequal time periods.. I can define duration of each period as
a) time between start and end of the period.
b) interval between mean time of the period t1 and mean period t0.
c) interval between mean (or median) resigthing date at periods t1 and t0.
b) and c) look more logical though I don't know what is a normal to go?

2. With MS mark-resight is it possible (complicated) to adjust McClintock WinBUGs and/or R code (http://www.stat.colostate.edu/~jah/software/) to the multistrata version? I love the possibility to account for heterogeneity?
Thanks,
Eldar

Hi Eldar,

For your first question, if you're referring to the length of time between the distinct primary sampling periods being different as a result of the sampling periods themselves being of different length, then check out Chapter 4 of 'The Book'.

For your second question, if one could obtain BUGS code for the traditional mark-recapture multi-state robust design (surely someone has written this?), then it would be (relatively) easy to modify it for mark-resight. One would just need to replace the mark-recapture abundance likelihood with the mark-resight abundance likelihood.

Cheers,
Brett