Initial results from using a GLMM to standardise the commercial CPUE data for Namibian hake, and an update of the GLM analyses to incorporate the extra data now available for 2004 and 2005
Working Paper
2006
Permanent link to this Item
Authors
Journal Title
Link to Journal
Journal ISSN
Volume Title
Publisher
Publisher
University of Cape Town
Faculty
License
Series
Abstract
Previous analyses to standardise the commercial CPUE hake data have encountered problems related to the presence of a significant year-vessel interaction effect in the GLM analyses. These include the fact that the CPUE data do not represent a balanced design and also that the number of parameters to be estimated exceeds the capabilities of statistical packages available. Furthermore, the presence of latitude or depth interactions with year in the GLM analyses complicates the standardisation of the CPUE series as integration over area becomes necessary, and procedures are needed to deal with missing cells. GLMM methodology is applied to overcome these problems. The conventional GLM analysis to standardise the CPUE for hake is applied to updated data for the period 1992 to 2005. The impact of a grid sorter on vessels is taken into account. The GLM standardised CPUE index shows a downward trend from 1999. An upward trend is evident from 2002 until 2004, but there is a 17% decrease from 2004 to 2005. Grid sorters are estimated to decrease catching efficiency by some 3%. GLMM standardised CPUE series show similar trends to those for the GLM, though the GLMM standardised series reflect a slightly larger decrease over the full period considered. The GLMM that includes all year interactions (i.e. including that with vessel) as random effects is selected as the best model of those considered in terms of both the AIC criterion and deviance analyses.
Description
Reference:
Brandão, A., & Butterworth, D. S. (2006). Initial results from using a GLMM to standardise the commercial CPUE data for Namibian hake, and an update of the GLM analyses to incorporate the extra data now available for 2004 and 2005.