.
My study consists of 12 streams with controls and 2 treatments, defaunation and a location (headwater vs. mainstem tributary) treatment, and 6 primary sampling dates (occasion 1 is the defaunation treatment). Three-pass removal electrofishing was conducted at each site while the population was closed (a removal-type robust design). All fishes were individually marked throughout the study (excepting those removed from the site at date 1, of course) to compare survival and immigration among treatments.
My first question is, if I have already arrived at a best model using AIC for estimating 'p' in a Huggins closed capture model, is it incorrect to use these values of 'p' to fix 'p' in subsequent CJS or Pradel models? In other words, do I just need break down and create entirely new capture histories for a Robust Design model provided in MARK or can I simply use 'p' from the initial closed capture models to fix 'p' in the open population parameters?
My second question is whether it is correct or possible to compute averages of the derived N-hat values from Huggins estimator by treatment groups (i.e. via delta-method?) and compare confidence intervals? The initial Huggins analyses utilized each sample point (every site and primary date =72) as groups in the capture history and as a starting point model, thus this is the same structure in which the N-hats are estimated even though a much reduced models was selected as the best fitting model for estimating 'p'.
Thanks for any suggestions,
Roy