covariates to model trap effects

[fpc]

Hello all,

I have a question regarding modeling trap response on detection prob. in the CJS format. Is it possible to add a dummy covariate that expresses previous detections to account for this? For example with:

Release -> recap1 - > recap2

Would S2 (survival between recap1 and recap2) be:

B0 + B1 * (a dummy variable expressing detection at recap1)?

Where I'm trying to get to is the Trap response in capture probabilities for CJS discussed in Williams et al. (2002) pg 427-428 and am lost as how to program this in MARK.

Thanks for any tips!

JT

[gwhite]

JT:

You've got the right idea -- but the covariate has to be a time-varying individual covariate. So, define a covariate that is 1 if captured on the previous occasion, and 0 if not captured on the previous occasion. This covariate is easy to create by just using the encounter history -- breaking it apart into covariate values. Suppose you name these covariates C1, C2, C3, ...

Then the section of the design matrix for p that you are interested in looks like:
1 C1
1 C2
1 C3
1 C4
...

You have the time varying covariate corresponding to the appropriate p.

This approach in MARK is equivalent to the paper Sandland, R. L. and G. P. Kirkwood. 1981. Estimation of Survival in Marked Populations with Possibly Dependent Sighting Probabilities. Biometrika 68:531–541.

You can also set up other time varying covariates such as the number of previous captures, or lag the above covariates at longer intervals.

I used this trick on desert tortoise data where the animal urinates when captured, losing a lot of water -- obviously not good for a desert species. I was interested in whether survival was affected for that interval. I don't remember the results, just the trick!

Gary