formatting individual covariates for age classes

[jbauder]

I have a 14 year mark-recapture data set for western rattlesnakes that I am using to obtain summer and winter survival estimates. Rattlesnakes were captured, marked, or recaptured at their dens in the spring and fall. Although this appears to be a straightforward CJS analysis, I would like to estimate separate survival estimates for neonates, juveniles, and adults. We are able to accurately assign these age classes whenever we capture a snake so some individuals were marked as neonates and recaptured as juveniles and/or adults or captured as juveniles and recaptured as adults. I am not clear on how to code categorical individual covariates like this in a CJS framework in a way that allows for separate survival estimates for each age class. Because the state of the snake (its age classes) changes during the course of the study, would a multi-state model be more appropriate?

[jlaake]

Categorical individual covariates are most easily handled with groups. Create 3 groups for your 3 age classes. No need for multistate because the transitions are known.

--jeff

[jbauder]

Thanks for the prompt response. It looks like I can set up the PIMs exactly as shown in the Gentle Introduction, Age and cohort model section (pg. 7-22 to 7-29). In my case, would my three groups be 1) marked as neonates, 2) marked as juveniles, and 3) marked as adults?
I have another question about setting the interval times. In our sampling design, we sampled our dens during the spring and during the fall. I will try to calculate seasonal survival rates (summer and winter) for a sub-set of the data, but for some years we did not visit the dens in the fall or had no recaptures in one season. So I would like to estimate annual survival rates for the entire data set. Is there any special way to treat the sampling intervals, since I am pooling two sampling events (spring and fall captures) that are separated by five months to represent a single, annual sampling event? Or can I still set my sampling intervals to one? I would like to obtain annual survival estimates in either case, but I wanted to see if I need to be aware of anything when interpreting annual survival estimates with my type of sampling design.
The neonates also may create an issue because an individual snake is only a neonate for about 8 months, from it's birth in the fall to emergence next spring, after which it becomes a juvenile. So while juveniles and adults can be in their respective age group for a year or more, neonates are only neonates for less than a year. Do I need to adjust for this in any way or can I set up my data and PIMs as I would with three normal groups and interpret my annual survival estimates as annual survival estimates? Would the "annual" survival estimates I obtain for neonates be valid and just refer to survival from birth in fall to spring emergence, instead of for an entire year?

Thank you very much for your help
Javan

[jbauder]

I wanted to re-post my question about how to treat multiple sampling events within a year when the goal of the analysis is annual survival estimates. My data set has two sampling events per year, one in spring and one in fall. I am interested in annual survival estimates and am wondering if I combine the data from the spring and fall sampling event to obtain "one event" per year and then analyze this combined data set as though there were one sampling event per year? I have found I do not have the data to analyze survival estimates for the periods between sampling events (i.e., summer and winter survival estimates).
Any insight into this question would be greatly appreciated!

[jlaake]

I would not merge them into one. I'd leave them as separate and include the time intervals between occasions and then if you don't have enough data to estimate survival separately for winter and spring you can use the same rate for the two periods. When you didn't sample in specific occasions you can set p=0.

--jeff