shared PIM question

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

shared PIM question

Postby jcohen » Mon Sep 16, 2019 9:09 am

I emailed Jeff Laake with the following question and he answered but asked me to move the topic here. I have pasted in the email thread and added a followup question below it.

My original question to Jeff:

"My student wants to see how a particular type of tag effects bird detection (and survival, although hopefully it doesn’t affect the latter). Birds were banded with a metal leg band if she caught them in a mist net (capture method 1) but then a couple of years later if they were recaptured, they received a PIT tag on the opposite leg from the metal band, after which they could be detected at nests with an RFID reader (capture method 2).

So…we have a potential multistrata model, with the states being A=“band only” or B=“band and pit tag”. I feel that in this case, PsiAB would equal pA (starting in the years she began using the PIT tags), because birds all received a PIT tag if they were recaptured. Does that seem correct to you? So that can be accomplished in MARK using the design matrices. But I am not sure we can do it in RMark (share the PIMS between pA and PsiAB). If this helps…we can use a logit link for Psi because there are only 2 states. Also PsiBA = 0."

Jeff's answer:

"Short answer is no, you cannot do that. But that is not how I would do it. Instead, I suggest that you use loss on capture approach. Let's say you had a bird you released with type A on occasion 1 and then saw it on occasion 3, put on B and then saw again on occasion 5. Your data would look like

10100 -1 A
00101 1 AB

The -1 tells MARK that the individual was not released. Other than tag type any individual covariates would be the same because they are the same critter. You would then use tag type A and AB as a group covariate for p or Phi in a CJS model."

And now my followup question:

If we use this approach, can we use AIC to compare the model Phi(g)p(g) to Phi(.)p(.)? My concern is that many animals are represented twice in the dataset, and I don't want to pseudoreplicate for the null model.
jcohen
 
Posts: 3
Joined: Thu Nov 07, 2013 8:50 am

Re: shared PIM question

Postby jlaake » Mon Sep 16, 2019 11:55 am

This is not pseudo-replication. In the likelihood each release of an animal is treated as independent unit which is equivalent to splitting up the capture history in this fashion.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA


Return to RMark

Who is online

Users browsing this forum: No registered users and 13 guests