Mark resight with group covariate

[mjordan]

I have a question about running a robust design Poisson log-normal mixed effects mark-resight model with a group covariate. We used camera traps to estimate density and survival of a population of fishers in California. Because we can't determine the sex of unmarked animals, we did not run any models with sex as a covariate. One of the reviewers of our manuscript suggested that we could run models that estimate survival and temporary emigration because those are based just on the marked individuals. That makes intuitive sense, but I can't see how to format the data without assigning the unmarked animals to a particular group. Am I missing something?

[gwhite]

If you were to code the marked data as 2 separate groups, then you are right that you would have to classify the unmarked animals according to sex. However, an alternative is to use an individual covariate of sex for the marked animals, coded (1,0) or (1,-1). The mark-resight estimators provide an estimate of the mean capture probability from the marked animals, and the marked animals are assumed to be representative of the unmarked animals. So, applying the mean capture probability of the marked animals where sex is known is appropriate because the unmarked animals are assumed to have the same sex ratio, and hence the same mean capture probability.

But, its up to you to be sure you meet these assumptions!

Gary

[mjordan]

That makes perfect sense. Thanks for the prompt reply!

-Mark