Robust design when Adding animals--help!

[brunner]

I'm relatively new to mark, so I hope this isn't too niave, but I am pretty stumped. In basic terms, I am wondering how to deal with animals that I am adding to a study site, and secondly, whether or not the timing of those additions matters.

The study involves measuring the densities of mice in forest fragments using a robust design--two days of trapping every three weeks--and on some of the fragments also adding more mice (captured elsewhere) every three weeks. It is the 'addition' sites that have me confused. It seems like these added mice will really alter the estimates of phi and p. Is there a way to fix these parameters at zero for the trapping sessions when they are not present?

Secondly is the matter of timing. For logistical reasons it would be best to add the mice the week before each set of two trapping sessions. But then these animals would be entering the model two-thirds of the way through the larger interval, and so I have no idea how to either code them or keep them from biasing survival estimates. Am I understanding the problem correctly? Do we need to make the additions at the same time as the trapping sessions?

I'd very much appreciate any help or points in the right direction. Thanks in advance!
-Jesse Brunner

[brunner]

I thought I would try one more time and see if anyone could point me in the right direction. (Sorry for the repetition)

I am unsure about how to treat animals that I am adding to a trapping grid at various points in time throughout a year (roughly every three weeks). I want to estimate the size of the entire population (animals present before and animals being added), so I don't want MARK to treat these added animals as if they were there before but just not caught.

Again, I'd appreciate any and all help or comments! Thanks!

[ganghis]

Probably what you want to do is to ignore the initial 'release' of these animals and to instead condition on the first reencounter. Since the robust design conditions on this first reencounter you should be fine. This approach simply requires editing the capture histories of released individuals so that the first '1' occurs when they are caught during one of the secondary sampling periods.

If for some reason you add new animals to the population in a period where you assume the population is closed (i.e. in between secondary sampling periods) and still want to estimate population size, it might be a good idea to enter these animals in as a second group so that you can constrain their capture probabilities to be 0 when they weren't in the population. The 'grand mean' population estimate would then be the sum of the two groups estimates with variance obtainable by the delta method.

[brunner]

Thanks for the advice! To be honest, I'm not entirely sure how the releases will be done in relation to the secondary encounters--eventually logistics, weather, etc. take precedence--so thank you for addressing both posibilities. But this helps give me a target to shoot for.
--Jesse