Multi-strata / robust design questions

[dglover]

Hi!

I am a relatively new user of MARK and I was hoping to get some help. I have four years of mark-recapture data of adult fish (only one group) from a very large lake. Tagging was conducted in a hierarchical design (secondary sessions within a primary) in four different states. In each state, tagging was done at a number of reference sites. Although I have a pretty large data set (approximately 60,000 tagged fish), I suspect I have a pretty low capture/ recapture probability (about 6,000 re-encountered). Additionally, “encounters” came from 1) the agency recaptures during the tagging operations, 2) angler recoveries, and 3) angler resightings. I am interested in determining movement transitions at the very least between the four states, survival rates and local population abundances. I have several questions which are listed below:

1) Concerning the different types of recaptures – It seems like I have an ideal situation to use a combination of Barker’s robust design and a multistrata model. Does anybody know if such a model exists or if there are any plans to incorporate them together? What would be the next best option if this model doesn’t exist? Most recaptures came from the tagging agencies, but not using the angler recoveries would also cut out a lot of encounters. Could I use the “Multi-strata – Live and Dead Enc.” and set the time intervals to correspond with a robust design? Should I just use different models to get at each question separately?

2) Although they were fairly close in time, not all tagging was conducted on the exact same dates for each of the states. Not all states had the same number of sampling dates either. Is there a way around this and how can I incorporate this into my encounter history? If the maximum number of sampling occasions within a year is six, can I just put a string of zeros in the site that was not sampled at that time? Would I then have to fix parameters for that site to zero or something similar?

3) I know that the number of parameters estimated with the multistrata models increases pretty quickly with the number of sites. Is there any rule on the maximum number of sites you should use with the amount of data you have?

Thanks in advance!

Dave