The data structure and models considered here can all be thought of as special cases of CJS models, which have in common that we are tracking marked animals through time. These are known as "conditional" models, in that the statistical likelihood is conditional on the initial release of marked animals, which are then re-encountered. By contrast, "unconditional" models also include captures of unmarked animals in the likelihood, needed for estimation of abundance and recruitment, as we will see later.
These conditional models all seek to estimate survival between encounter occasions, which may in turn be modeled as group-specific, time-specific, or a function of individual or time covariates. Depending on the specific data structure, additional parameters may need to be estimated.
CJS: Re-encounter probabilities assumed to be <1 and are estimated; denoted as p (recapture probability). Data are initial release occasion of each animal and a history of live re-encounters.
Dead recovery: Re-encounter probabilities assumed to be <1 and are estimated; denoted as r (reporting) or f (recovery) probabilities, depending on model parameterization. Data are initial release occasion of each animal and occasion of dead recovery (if any)
Joint live-dead: Re-encounter probabilities assumed to be <1 and are estimated. Data are initial release occasion of each animal, live recaptures or resighting, and occasion of dead recovery (if any) . Parameters include apparent survival (Phi), reporting (r), recapture or resighting (p) probabilities. Depending on data structure and assumptions, additional parameters such as fidelity (permanent emigration) may be estimated.
Next: CJS review