Midwest Fish & Wildlife Conference 2016

AUTHORS: James R. Bence*, Michigan State University

ABSTRACT: For the purposes of this presentation, a state-space model is one where the probability distribution of the system state at the next time step is determined by the system state at the current time step, and transition rules, and where system characteristics are observed with error. The estimation task is estimate parameters for the transition rules, the initial system state, and process errors. Given this definition, most recent statistically-based fishery stock assessments are state-space models. For fishery stock assessment, interest is in both identifying and parameterizing the transition rules, and in determining the likely system states over the observation period, particularly stock size and mortality rates at the end of the time-series. The underlying transition rules are critical for supporting evaluation of management strategies, whereas estimation of the current state is needed for applying a management strategy (e.g., setting a limit on annual harvest). A diversity of estimation approaches are currently applied in fishery stock assessment, including a "mixed-effect" approach where process errors are treated as random effects, penalized likelihood (from a Bayesian perspective highest posterior density), and full Bayesian hierarchical approaches where process errors are drawn from distributions, and inferences about the parameters of those distributions are also made. The strengths and weaknesses of these various approaches will be reviewed.