Accounting for individuals "changing groups"

Forum for discussion of general questions related to study design and/or analysis of existing data - software neutral.

Accounting for individuals "changing groups"

Postby Rebekah Ry » Fri Jul 31, 2020 2:18 pm

Hello,

I have searched the forum and haven't found an answer to the question I have, so hopefully this isn't a repeat of what someone has already posted.

In a nutshell, I work with a social species of bird. I color band individuals (both parents and young) at nestboxes, and I focal monitor these "family groups" year-round. Often juveniles will remain with their parents for an extended period of time, and frequently these juveniles will set up territories near their father's territory, ultimately creating kin-structured neighborhoods over time. Established breeding pairs have high site fidelity and will nest in the same nestbox (or a nearby nestbox) year after year. I have 7 years of this type of data for over 200 "family groups" and over 1000 individuals.

I hypothesize that individuals within "family groups" have a higher apparent survival than individuals not residing in a family group, and I want to take this into account in my CJS model. I have witnessed relatives sharing food resources and engaging in fewer aggressive interactions with conspecifics (another chapter of my dissertation). However, I have no idea how to code for this in RMark. Obviously from year to year, "family groups" change. For example, in 2013, male and female fledge young from nestbox 1-3. Then in 2014, the same male and female use nestbox 1-4 and their son from the previous year is now occupying nestbox 1-3. In 2015, male is with a new mate in nestbox 1-5, and son is still around using nestbox 1-3. And so on and so forth.

How do I take into account these time-transitioning "family groups?" For other models, I have used "family group" as a random factor, but I know that in RMARK (implemented through Mark), its different. I also know that I eventually am going to run a multistate model for apparent survival taking into account changing social status, but for my CJS model, how do I code this? Not all individuals change "family groups," and most of my individuals disperse and are never seen again.

Hopefully that made sense. Any help would be amazing. I love it that my birds are complicated, but it makes for some rough modeling!

All the best,
Rebekah
Rebekah Ry
 
Posts: 13
Joined: Wed Jul 15, 2020 6:46 pm

Re: Accounting for individuals "changing groups"

Postby jlaake » Fri Jul 31, 2020 9:52 pm

If you knew whenever a bird changed groups you could use loss on capture approach and move them to new group. But you aren't always going to know that presumably and it is not clear to me how you are treating the dispersers in that case. I think the only way you could do that is in a multistate model and even then do you see birds that aren't in a family group.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: Accounting for individuals "changing groups"

Postby B.K. Sandercock » Mon Aug 03, 2020 3:46 am

Hi Rebekah:

I agree that a multistate model might be a better approach than a CJS model for your dataset with dynamic categorical states. You could start with three possible states: juvenile neighbour (J), breeding adults without neighbours (A), and breeding adults in a family group (F). The multistate model would give you state-specific estimates of apparent survival, which would partly address your question about survival advantages of social groups. If apparent survival is ranked Phi(J) < Phi(A) < Phi(F) then it could be due to differences in true survival or to site fidelity. The multistate model would also give you estimates of the transition rates among the states which might be of interest too, which you could model vs. environmental covariates. If there is no regression in your social system, you could fix the transition parameter for Psi(A to J) to zero. And if data are sparse, you could collapse the model to two states for juveniles and adults. For a published example of a 3-state model, take a look at the following article:
Schaub, M., Jakober, H., and Stauber, W. (2011). Demographic response to environmental variation in breeding stopover and non-breeding areas in a migratory passerine. Oecologia 167: 445–459.
Good luck with your analyses.

Regards, Brett.
B.K. Sandercock
 
Posts: 48
Joined: Mon Jun 02, 2003 4:18 pm
Location: Norwegian Institute of Nature Research

Re: Accounting for individuals "changing groups"

Postby Rebekah Ry » Wed Aug 05, 2020 10:56 am

Hello Jeff and Brett,

Thank y'all so much for your input. It's awesome that these kinds of forums exist to help out us non-experts with difficult questions!

Y'all are right in that I need to use a multistate model. There isn't a solid way to "categorize" social state without using transitions. I will definitely look into the resource you shared, Brett (thanks!). However, if I wanted to estimate apparent survival of only AHY adults, not taking into account social state, I would be well off using a CJS, right? Would it even make sense to run this additional analysis if I am already running the multistate? Originally I had three analyses in the final chapter of my dissertation proposal: 1). Apparent survival of adults only (CJS), 2). Nesting success, recruitment, and pop growth rate (Kaplin-Meier, Pradel), and 3). Apparent survival taking into account social status (Multi-state).

Any feedback would be awesome. Thank you so much for your time and expertise!

Best,
Rebekah
Rebekah Ry
 
Posts: 13
Joined: Wed Jul 15, 2020 6:46 pm

Re: Accounting for individuals "changing groups"

Postby B.K. Sandercock » Fri Aug 07, 2020 7:21 am

Hi Rebekah:

You could recode your encounter histories to reduce the 3-state multistate model to separate analyses for juveniles and adults. The 3-state model would give you all of the parameters you are interested in but you’d need adequate data for estimation of all parameters. Example of encounter histories:
J0000000
J0A0AFF0
00AFF0F0

Juveniles only. For juveniles only, you could recode the encounter histories to a single-state of detections and nondetections, and censor records of birds marked as adults. Run an age-structured CJS model on the encounter histories to get age-specific apparent survival. This approach is often a good idea for vertebrate populations if you have a lot of individuals that are marked in their natal year but never seen again. Example histories:
10000000
10101110
00AFF0F0 (censored because no juvenile records).

Adults only multistate. You can also drop juveniles from the encounter histories by recoding records of juveniles as nondetections.
00000000 (censored because no adult records)
00A0AFF0
00AFF0F0
A 2-state multistate model for adults would address the questions that you are interested in about effects of neighbours on apparent survival of adults. Might start with this model but if the parameter estimates are noisy and hop around, are stuck at the boundaries of zero and one, or have poor precision, you could also run a CJS model on the adults

Adults only CJS. You can convert the multistate encounter histories for adults by recoding any detections as adults as ones.
00000000 (censored because no adult records)
00101110
00111010
The CJS model for adults will pool any differences between birds of different social status and will give you an average estimate of apparent survival of adults but with better precision than the multistate model. Here’s an example of a paper where we ran CJS and multistate models on the same set of encounter histories (Sandercock et al. 2000 Ecology, 81(5):1351–1370).

Regards, Brett
B.K. Sandercock
 
Posts: 48
Joined: Mon Jun 02, 2003 4:18 pm
Location: Norwegian Institute of Nature Research

Re: Accounting for individuals "changing groups"

Postby Rebekah Ry » Sat Aug 08, 2020 12:39 pm

Hey Brett,

Thank you so much for your lengthy and thorough response - this has definitely given me a lot to think about, and I am likely going to re-plan my analyses based on your suggestions. I like the idea of breaking the age groups into separate models, because you're spot-on - most of the birds I band are nestlings that disperse (or die) and are never recorded again. I don't want this to detract from the solid data I do have collected for adults.

Much appreciation for the help and resources!

Cheers!
- Rebekah
Rebekah Ry
 
Posts: 13
Joined: Wed Jul 15, 2020 6:46 pm


Return to analysis & design questions

Who is online

Users browsing this forum: No registered users and 15 guests