Structured additive distributional regression for analysing landings per unit effort in fisheries research

详细信息    查看全文

• 作者：Valeria Mamouridis^a ; ^{mamouridis@icm.csic.es} ; ^{valeria.mamouridis@gmail.com} ; Nadja Klein^b ; Thomas Kneib^b ; Carmen Cadarso Suarez^c ; Francesc Maynou^a
• 关键词：Landings per unit effort ; Red shrimp fishery ; North Atlantic Oscillation ; North-Western Mediterranean ; Generalised additive models for location ; Scale and shape ; Mixed models ; Markov chain Monte Carlo simulation
• 刊名：Mathematical Biosciences
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：283
• 期：Complete
• 页码：145-154
• 全文大小：1131 K
• 卷排序：283

文摘

Structured additive distributional regression models constitute a powerful tool to analyse determinants of and variability in fishery indices. The gamma distribution and the flexible estimation for both first and second order parameters of the distribution describe the LPUE patterns for the particular case as supported by results of the DIC, the quantile residual plots and MSEP. Results also show significant improvements when accounting for the second order parameter. The shape was related to many of the explanatory variables. Mixed models are particularly suitable for unobserved fishing units and for correlated observations in time series analysis, even if our observations represent the whole population of the studied fleet. Fixed effects models are incapable to include variables that are constant for a unit (vessel in this case) and so restrictive for understanding the unobserved heterogeneity.