Standardization of CPUE from Aleutian Islands Golden King Crab Fishery Observer Data
Our primary task is to standardize the catch-per-unit-effort (CPUE) of observer pot sample data to input to the Aleutian Islands golden king crab (GKC) assessment model (Siddeek et al., 2013). We presented different versions (improvements) of analysis of observer and fish ticket data using generaliz...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Text |
| Language: | English |
| Subjects: | |
| Online Access: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.396.5127 http://209.112.168.2/npfmc/PDFdocuments/membership/PlanTeam/Crab/September13/AIGKCCPUE.pdf |
| Summary: | Our primary task is to standardize the catch-per-unit-effort (CPUE) of observer pot sample data to input to the Aleutian Islands golden king crab (GKC) assessment model (Siddeek et al., 2013). We presented different versions (improvements) of analysis of observer and fish ticket data using generalized linear model (GLM) and generalize additive model (GAM) at the May and September 2012 Crab Plan Team (CPT) meetings. For the February 2013 crab model workshop, we switched back to GLM because it can handle any nonlinear situation either by converting the variables into categorical variables or by using cubic splines. Furthermore, GLM is based on sound statistical principles (Fox and Weisberg, 2011, Zuur et al., 2009). Following suggestions made at the February 2013 Crab Modeling Workshop, we redid the GLM analysis restricting it to observer pot sample legal size male data and dividing the time series into pre (1995/96– 2004/05) and post (2005/06–2010/11) rationalization periods. We treated the explanatory variables: Soak, Depth, and product of number of vessels and mean soak time, as numeric and used the piecewise-cubic spline to describe their functional forms in the GLM. For one scenario, we departed from lognormal and 1 binomial GLM families to the negative binomial family. Following workshop suggestion, we forced in the Soak variable (if not selected) to all final models. We computed yearly CPUE indices with confidence |
|---|