/HookandLineSurveyCatchAnalysis

Binary catch data analyzed at the hook level using MCMClogit() in the MCMCpack R package.

Primary LanguageR

Hook_and_Line_Survey_Catch_Analysis

Binary catch data analyzed at the hook level using a modified stepAIC() function from the MASS R package with a sample from the posterior distribution generated using MCMClogit() in the MCMCpack R package.

The modified stepAIC() adds a buffer to the AIC testing of the models being looped over. The new candidate model has to be better (smaller) than (AIC - buffer) of the current model for the break of the loop not to occur. So if the buffer = 1.95 (the default), then the new candidate model has to be almost 2 AIC units better for the AIC stepping to continue. If the buffer is significantly greater than 2 AIC units then the new model has to be a lot better than the current model for the AIC stepping to continue. Hook and Line Survey binary GLM models appear to need a value far greater than 2 AIC units to have parsimony with not too many variables and biologically dubious interactions terms. This need is not fully understood, but the hook level data does make the total N in the models relatively high and a large number of explanatory variables are considered. In practice the buffer works quite well once it is tuned. The buffer is a vector, with the first element being for main effects and the second element for interaction terms.

Note that the early years of the survey had less non-CCA (Cowcod Conservation Area) sites then the standard 121 unique sites used in the non-CCA part of the survey:

     year numSites
 1  2004       74
 2  2005       90
 3  2006       92
 4  2007       99
 5  2008      120
 6  2009      118
 7  2010      121
 8  2011      111
 9  2012      121
 10 2013      120
 11 2014      120
 12 2015      115
 13 2016      121
 14 2017      120
 15 2018      122
 16 2019      122

For species that are caught less frequently and predominately only on certain sites, these missing sites may impact the index in the early years.

Years with less sites than 121 after 2008 are due to weather issues. A 122'nd site was added starting in 2016 but is not productive and may be removed in the future.

Reference

Analysis of fishery-independent hook and line-based data for use in the stock assessment of bocaccio rockfish (Sebastes paucispinis)

 https://www.sciencedirect.com/science/article/pii/S0165783610002031