www.phidot.org

Hi everyone,

I am using RMark to estimate adult annual apparent survival for bluebirds. My ultimate goal is to not only estimate adult survival, but to determine which weather and climatic variables affect survival. I have CMR data for 2003-2018 (excluding 2011, when no trapping was done). Since I'm using CJS models, I tested the assumptions for transience (Test3.SR) and trap dependence (Test2.CT) in UCARE. Test3.SR was not significant, but Test2.CT was (specifically, the birds are trap-shy).

My understanding of how to deal with trap dependence is to create 2 classes of probability of capture, based on whether the bird was caught the previous occasion (TSM, or time since marking). The model should basically be Φt,pm*t. I looked through Appendix C of "Program MARK: A Gentle Introduction" and at previous posts on phidot.org for help on how to do this in RMark, but am a bit confused. I'm hoping someone on here can clarify this procedure.

Pages 77-81 of Appendix C offer explanations and example code for creating a trap dependence variable (td) in RMark. This is very straightforward, but the UCARE manual and another post on this forum (http://www.phidot.org/forum/viewtopic.php?f=1&t=808) mentions transforming the data using the "split for trap-dep analysis" feature in UCARE.

My question is, are these 2 steps independent means to the same end (i.e, do one or the other), or do I need to split for trap-dependence, THEN create the td variable in RMark? Or have I missed the mark (pun not intended) altogether? Any help clarifying this issue would be greatly appreciated. Please let me know if I didn't explain something clearly or more information is needed.

Thank you,
Sara

All I can really do is describe what Appendix C does. For occasion t+1, it splits the critters based on the value of occasion t (0 or 1). I can think of at least one other way to do it which would be a permanent split. Once caught it moves to a previously caught group (like a behavioral effect in closed models). But since you are talking about CJS which conditions on first release(capture) (the first 1), then you have to think about that. That would mean for all occasions that are modeled each animal would already be in the previously caught group so wouldn't make much sense if you used the first occasion. Unfortunately I don't know what the UCARE approach does, so I can't be any help there. Maybe someone else can chime in.

Thanks jlaake,

That solution makes sense to me, so that is how I decided to proceed. I'm still curious about the UCARE approach, but until I get more information, I'm sticking with the suggestion in Appendix C.