age AND time-since-marking model

[JoeEM]

Hello,

I am modeling bird captures in which all individuals are assigned an initial age of adult or subadult. We have many adult transients as well. To combine the two factors, I am testing a modified TSM model in which survival is estimated separately for juveniles, 1st capture adults, and subsequent capture (non-transient) adults. Here are sample PIMs (reduced to only 3 capture events for clarity):

subadults:
122
12
1

adults:
322
32
3

Is this a suitable approach, or can anyone recommend a better one? Thanks,

Joe Meisel

[cooch]

Hello,

I am modeling bird captures in which all individuals are assigned an initial age of adult or subadult. We have many adult transients as well. To combine the two factors, I am testing a modified TSM model in which survival is estimated separately for juveniles, 1st capture adults, and subsequent capture (non-transient) adults. Here are sample PIMs (reduced to only 3 capture events for clarity):

subadults:
122
12
1

adults:
322
32
3

Is this a suitable approach, or can anyone recommend a better one? Thanks,

Joe Meisel

What you have is fine, and pretty standard. But, be aware that you're assuming that adult survival (i.e., off the diagonal) is equivalent regardless of the age at which individuals were marked. While this seems reasonable, there are often good reasons to suspect that survival of adults marled as adults may be different from survival of adults marked as adults. Why? Well, consider the fundamental difference in your two groups. Marked as juveniles are natal recruits - you know where they came from. Marked as adults could be...anything! Controlling for 'transients' using a TSM model really only controls for one type of heterogneity (and a fairly limited type at best). There may in fact be other reasons to suspects that 'not all adults are really the same). I tend to recommend that testing equivalence of adult survival rates should be part of the model set. Your PIMs assume this is true. Why not simply let them be different, just to test this hypothesis?

So, basically something analogous to

Code: Select all

1 2 2
  1 2
    1


3 4 4
  3 4
    3


You can figure out the design matrix for this fairly easily, I suspect - section 8.1.2 gives some pretty good 'clues'.