I'm using a CJS model to evaluate salmon smolt survival. Many more individuals were PIT-tagged than was probably necessary for the first year of the evaluation, so I'm trying to provide some advice on whether and by how much to reduce the tagging rate (i.e., release group sizes) without compromising the ability to detect differences. In other words, I want to do a kind of power analysis.
In MARK, I realize this can be done using simulations in RELEASE (e.g., Mark book, appendix A) by varying the release sizes. However, I'm curious whether something similar could be done with the data boostrapping procedure in MARK, because my first thought, conceptually, was to do an analysis with resampling from the encounter histories (e.g., 1,000 bootstrap replicates randomly selecting 90% of the encounter history data at each replicate, 80% of the data, and so forth) and evaluate in terms of model selection (e.g, Mark book, Appendix G2). The appeal here is that it uses the actual encounter histories.
My question: with the bootstrapping procedure in MARK is it possible to specifiy a sampling rate from the encounter histories within each group? Or would I have to set up a resampling routine in RMark (I have limited experience with RMark but assume this would be possible).
Thank you!