In Part 8a, we treated times elapsed since release (initial cohort date) as "age" and created a series of models under this assumption. However, unless a release cohort is known to all be individuals of a specific age (as would be the case if, say , we were banding only nestlings or tagging spawn) we have created a situation where the "cohorts" are mixtures of animals of different ages.
A true age-specific analysis in more general circumstances requires us to
Be able to identify the age (or age class) of all animals when first captured.
Tag at least some individuals in each age class, each capture occasion.
We just considered the Serin 2-age example, where male and female birds were ringed in 2 age classes: SY (subadult) and ASY (adult). I have reformatted these data in the attached script to ignore sex but to consider birds as having been initially ringed in one of 4 ages classes:
HY (hatching year, so age =0 in initial year of release)
SY (second year, age =1 in initial year)
3Y (third year, age =2 in initial year)
ASY (after third year, age =3+ in initial year)
The code modifies the ddl to define age.now as 0 – 3, where 3 represents age 3+ birds. (note that now p *is* age dependent since birds can be recaptured as SY, 3Y, or ASY)
Confirm (by displaying the ddl) that the 'age.now' variable is correct for each year of the study
Re-analyze the Serin data under this age structure. Construct models with age.now, time, age.now*time, age.now+time, and no effects (constant Phi) on survival and recapture probability. Produce model averages estimates of the real parameters time and age-specific Phi and plot by age over time