by murray.efford » Tue Dec 20, 2011 2:48 pm
Miranda
The package 'secr' is currently for closed populations; you should use RMark/MARK for survival analyses. However, (i) the latest version of secr (2.3.1 became available on CRAN a few hours ago) has a new function RMarkInput that makes it easy to flip data from `secr' to RMark (see ?RMarkInput) and (ii) secr 2.3.1 has an undocumented 'death-only' (survival but no recruitment) open population model that will be extended and documented next year. In the meantime, insider knowledge is needed!
Using nonspatial methods has a slight cost: CJS estimates of survival probability (as implemented in MARK) will be somewhat biased by the heterogeneous individual capture probabilities that usually result from spatial sampling. This is usually assumed to be negligible, but may not be. So, ultimately, 'spatial' survival estimation that models the spatial individual heterogeneity is a good idea, but the software is not quite there.
If you are not primarily interested in survival but rather in population trend then 'secr' will do the job, as it can fit a density trend across multiple closed population 'sessions'.
Murray