Premium
Catch‐Rate Standardization for Yellow Perch in Lake Erie: A Comparison of the Spatial Generalized Linear Model and the Generalized Additive Model
Author(s) -
Yu Hao,
Jiao Yan,
Winter Andreas
Publication year - 2011
Publication title -
transactions of the american fisheries society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.696
H-Index - 86
eISSN - 1548-8659
pISSN - 0002-8487
DOI - 10.1080/00028487.2011.599258
Subject(s) - generalized linear model , generalized additive model , statistics , autocorrelation , goodness of fit , econometrics , general linear model , spatial analysis , mathematics , stock assessment , generalized linear mixed model , linear model , ecology , biology , fishing
Various statistical models have been used to standardize catch rates of fish populations for stock assessment. The generalized linear model (GLM) is one of the most commonly used approaches. However, response variables in fisheries data are often spatially autocorrelated or show a nonlinear relationship with explanatory variables, thus violating the underlying assumption of the GLM. The GLM with spatial autocorrelation (s‐GLM) and the generalized additive model (GAM) may be better suited to dealing with autocorrelated variables and nonlinear relationships, respectively. In this study, catch rates of yellow perch Perca flavescens from a fishery‐independent gill‐net survey in Lake Erie during 1990–2003 were estimated by the s‐GLM and GAM. Set duration, gear depth, temperature at gear depth, dissolved oxygen concentration, water transparency, and latitude were selected as significant explanatory variables for these models. By comparing the goodness of fit, reduction of spatial autocorrelation, and prediction errors from cross validation, we found that the GAM had the best goodness of fit and lowest prediction errors but the s‐GLM resulted in the greatest reduction in spatial autocorrelation for catch‐rate standardization of yellow perch in Lake Erie. The GAM is more suitable for modeling the nonlinear relationships between species characteristics and explanatory variables, and the s‐GLM is preferable for dealing with spatially autocorrelated residuals. Application of the GAM or s‐GLM is determined by the species distribution pattern and available explanatory variables.