MS live dead with unobservable states including fidelity
Posted: Thu Jan 24, 2013 2:44 pmHi,
I have a question pertaining the use of Live-dead data. I am working on a owl species where many (most) individuals are marked at the nest as newborns. Some of them may be recovered dead, some may be found alive during some breeding season later on in time, and some, I presume, leave the area and cannot be retrapped alive anymore. I have to add that live recaptures are usually at the nest (almost exclusively). A normal LD model provides me with the well known survival, recapture, fidelity and recovery rates, whose estimates make sense given external information from the literature. I should perhaps mention that Fidelity is modelled as age dependent (2 classes) contrasting juveniles and older age classes (with adult fidelity close to 1). I also modelled the recapture rates as age dependent (again 2 age classes), and discovered a much much lower recapture rate for juveniles than older ages. This made me think that there perhaps might be an issue of delayed recruitment, which should make juveniles that have not recruited into the breeding segment of the population yet, be less (or not) capturable. I thus attempted to use Barker's MS Live dead model, and add an unobservable state of non-breeders, subject to the constraint that all individuals still alive by the end of their third year become breeders, and subject to recapture rates (but not recovery rates) for non-breeders fixed at 0. This is reasonable assumption in the case of the focal species. The models converged well (judging by both SANN and MCMC), however my juvenile survival rate only about half of that provided in the classical Burnham's model. So so far, the classical model provides me fidelity, and does not directly provide me recruitment probability, and Barker's MSLD might provide me with recruitment, but apparently cannot account for permanent emigration. The next step I took is to add another unobservable state that I called "emigrated". I fixed recapture rates at zero, allowed only non-breeders in the first year to emigrate, and assumed equal recovery rates for all three states. I guess my question is: does anyone have experience with a similar analysis? Are the parameters theoretically estimable? Or am I better of taking the ratio of juvenile to adult recapture rates as an approximation? I am asking these questions before moving on to a complicated simulation study.
Actually, my case is more complicated, because I would like to model survival as function of fledging condition, but would also like to exclude the possibility of higher/lower dispersal rates given condition, and recruitment rates which also may be a function of fledging condition. Ultimately, I am interested in cost of breeding, although I realize I may be pushing it too far.
Thank you in advance for any thoughts and comments.
Best regards
Caspar
I have a question pertaining the use of Live-dead data. I am working on a owl species where many (most) individuals are marked at the nest as newborns. Some of them may be recovered dead, some may be found alive during some breeding season later on in time, and some, I presume, leave the area and cannot be retrapped alive anymore. I have to add that live recaptures are usually at the nest (almost exclusively). A normal LD model provides me with the well known survival, recapture, fidelity and recovery rates, whose estimates make sense given external information from the literature. I should perhaps mention that Fidelity is modelled as age dependent (2 classes) contrasting juveniles and older age classes (with adult fidelity close to 1). I also modelled the recapture rates as age dependent (again 2 age classes), and discovered a much much lower recapture rate for juveniles than older ages. This made me think that there perhaps might be an issue of delayed recruitment, which should make juveniles that have not recruited into the breeding segment of the population yet, be less (or not) capturable. I thus attempted to use Barker's MS Live dead model, and add an unobservable state of non-breeders, subject to the constraint that all individuals still alive by the end of their third year become breeders, and subject to recapture rates (but not recovery rates) for non-breeders fixed at 0. This is reasonable assumption in the case of the focal species. The models converged well (judging by both SANN and MCMC), however my juvenile survival rate only about half of that provided in the classical Burnham's model. So so far, the classical model provides me fidelity, and does not directly provide me recruitment probability, and Barker's MSLD might provide me with recruitment, but apparently cannot account for permanent emigration. The next step I took is to add another unobservable state that I called "emigrated". I fixed recapture rates at zero, allowed only non-breeders in the first year to emigrate, and assumed equal recovery rates for all three states. I guess my question is: does anyone have experience with a similar analysis? Are the parameters theoretically estimable? Or am I better of taking the ratio of juvenile to adult recapture rates as an approximation? I am asking these questions before moving on to a complicated simulation study.
Actually, my case is more complicated, because I would like to model survival as function of fledging condition, but would also like to exclude the possibility of higher/lower dispersal rates given condition, and recruitment rates which also may be a function of fledging condition. Ultimately, I am interested in cost of breeding, although I realize I may be pushing it too far.
Thank you in advance for any thoughts and comments.
Best regards
Caspar