Age Structure in PIMs?

[adamdillon]

Murray,

I'm having a bit of trouble conceptually trying to address age structure in my data and I was hoping to get some guidance.

I'm trying to estimate the density of foxes (both pups and adults) with a 7 year data set. Currently I have pups and adults set up in separate sessions, the problem is I would like some of the parameters to be shared between these two sessions.

The adult session consists of foxes that were first caught as adults and therefore are adults for all the years they were trapped. All animals in the pup session are considered pups the first year they are caught and then considered adults in all subsequent years. The current way my data is set up the adult and pup sessions are analyzed completely separately and what I need is for the 1st year that animals in the pup session were captured, to be analyzed separately from the 2nd+ plus years these same "pups" were captured (since from year 2 onward they are considered adults) along with all of the animals in the adult session . This would be analogous to creating an aged PIM structure in MARK, where the estimated parameters for the first year's capture of pups is different from the estimated parameters for the 2nd+ year's capture of pups, which in turn should be equal to the corresponding estimated parameters in the adults.

I am trying to estimate yearly Density for pups and adults and therefore I'm using the full ML. I hypothesize that pups will have a different overall density and a different capture probability than adults, but currently the way my data is set up this issue is confounded because the data in the 2nd+ years in the pup session are being analyzed as pups and they should be lumped into the adult session.

Hopefully this is slightly clearer than mud, but if not, let me know and I can try and do better at explaining my situation. Thanks.

Adam

[adamdillon]

So I think my mind is starting to turn to mush a bit.
I think transiting my data from MARK to SECR got me confused.

Correct me if I'm wrong but in MARK the data is entered by individual, therefore if an animal is originally captured as a pup it will always be in the pup "group" and therefore you need to create the cohort structure in the PIMs in order to transition the pup parameters to the adult parameters.

but in SECR the data is entered by capture occasion, therefore when a pup is captured in its first year, I can simply place it in the pup "group" but when that same animal is captured in the following year, I can simply place it in the adult "group". This would separate the pup and adult parameters, allowing me to estimates the parameters separately or test if there is differences between these groups by fitting two separate models.

If I'm way off base let me know.
Thanks
Adam

[murray.efford]

Hi Adam
Just catching up after a few days away...

I think you figured it out. Because 'secr' does not link individual capture histories across 'sessions', there is no (coding) problem when individuals change groups or sessions. Be warned that we are cheating a bit here - the model pretends sessions are independent when they strictly are not. For trend models this means the SE of trend across sessions is underestimated.

A fully spatial open model is what's needed, and although I have a lot of this together it's not ready to go out. That should have an 'ageclass' predictor variable that would meet your need.

On the side - I imagine you could achieve what you are currently doing with the full likelihood by the conditional likelihood route, with post-hoc grouping of animals (the groups argument of derived()).

Murray