Band recovery -> Multistate; Short sampling season required?

[claudiapenaloza]

I'm putting together a multistate model for band recovery data (i.e., one live banding encounter and one dead encounter or recovery) to try to elucidate migratory behavior in American black ducks.

As far as I understand, in "normal" live-dead multistate models (where you could have more than one live encounter in an individual's capture history), live encounters are only recorded during specified sampling periods, whereas recoveries can occur at any time of the year.

Given that I will only have one live encounter per individual, do I still need to restrict live encounters to relatively "short" sampling periods (the assumption of equal probability of survival between individuals will not be violated if there is only one live and one dead encounter, correct?) or can I use all the banding data that I have (which has bi-annual peaks but occurs throughout the year)?

Thanks for any/all help,
Cheers,
claudia

[claudiapenaloza]

I talked to Bill Kendall about this... and yes, I still need to have relatively "short" sampling periods because these captures/recaptures will still be condensed into only one primary period, as will the recoveries (to the intervals). If, for example, my live sampling period is 6 months long with a 6 month interval between samples and we recover two individuals right before the next sampling period, but one of them had been banded at the beginning of the previous sampling period and the other had been banded toward the end, then the first individual had to have survived an additional 6 months before it was recovered. Therefore, the assumption of equal probability of survival would still be violated with very long sampling periods, even if there is only one live and one dead encounter.

Please correct me if I'm completely off here.
Thanks,
cp