diff --git a/9control.tex b/9control.tex index 1b8a66c0..9fd5f8c8 100644 --- a/9control.tex +++ b/9control.tex @@ -2444,7 +2444,7 @@ \subsubsection{Predator Fleet Mortality} There are a range of controls for this feature. The deviations can be applied to either the age selectivity or the length selectivity over a specified range of ages (or lengths) and years. The variance of the devs can be the same for all ages, or can be age-specific up to a specified age, then constant variance for all older ages that are within the 2DAR range. Ages outside the 2DAR range revert to the parametric selectivity. The user can select a range of years for which devs will be created and can specify whether selectivity for years outside this range revert to the parametric selectivity or continue with the semi-parametric from the terminal year with devs; thus devs could be defined for just one year, the first year, then mirrored for all subsequent years. Finally, the controls allow for there to be a user-specified level of auto-correlation along the age dimension separate from a user-specified level of auto-correlation in the year dimension. -When using this option for age-based selectivity, if there are not too many ages, a good choice for the baseline selectivity might be random walk selectivity (pattern 17) because it provides the most flexibility, allowing the 2DAR deviations to be used only for the annual deviations around this baseline rather than the account for misspecification of the underlying functional form. Otherwise, a simple parametric selectivity form like logistic or exponential logistic would be a reasonable choice. The 2DAR option has not yet been explored adequately to provide guidance on best practices. +The 2DAR option has not yet been explored adequately to provide guidance on best practices. However, a preliminary guidance includes the following. When using this option for age-based selectivity, if there are not too many ages, a good choice for the baseline selectivity might be random walk selectivity (pattern 17) because it provides the most flexibility, allowing the 2DAR deviations to be used only for the annual deviations around this baseline rather than the account for misspecification of the underlying functional form. Otherwise, a simple parametric selectivity form like logistic or exponential logistic would be a reasonable choice. With regard to the variability, never estimate the sigma parameters as estimates will biased toward zero. Setting the sigma near 1.0 is advised as an initial default unless the user explicitly determines superior performance of a different value. \begin{tabular}{p{2cm} p{14cm}} \hline @@ -2465,19 +2465,18 @@ \subsubsection{Predator Fleet Mortality} \item Amin: First integer age (or population length bin index) with deviations. \item Amax: Last integer age (or population length bin index) with deviations. \item Sigma\_Amax: the last age (or population length bin index) for which a separate sigma should be read. The number of sigma parameter lines to be read will be Sigma\_Amax - Amin + 1. The feature allows for the expected situation in which annual variability in selectivity is higher for younger ages. For simplicity, only a single sigma parameter line needs to be read if Sigma\_Amax equals Amin, or if Sigma\_Amax < 0. - \item Use Rho (0/1): Use autocorrelation parameters. This invokes two parameter lines, first for year, second for age (or length). The rho parameter must be set as fixed; it is treated as a constant and cannot be estimated by SS3. + \item Use Rho (0): Use autocorrelation parameters. The rho feature of 2DAR is incorrectly implemented and should not be used (value must be 0). \item Len(1) / Age(2): to specify whether the deviations should be applied to length- or age-based selectivity. \item Phase: Phase to begin estimation of the deviation parameters. \item Before Range: How should selectivity be modeled in the years prior to Ymin? Available options are (0) apply no deviations, (1) use deviations from the first year with deviations (Ymin), and (3) use average across all years with deviations (Ymin to Ymax). \item After Range: Similar to Before Range but defines how selectivity should be modeled after Ymax. \end{enumerate} -Following each fleet-specific specification line are short parameter lines for the standard deviation of the devs (sigma\_selex) and, optionally, their autocorrelation (rho). +Following each fleet-specific specification line are short parameter lines for the standard deviation of the devs (sigma\_selex). \begin{enumerate} \item Short parameter line for sigma\_selex at Amin; - \item If Sigma\_Amax > Amin, then additional parameter lines for each age up to Sigma\_Amax; - \item If Use\_Rho = 1, then one parameter line for rho in the year direction and one parameter line for rho in the age (length) direction. These must be set to have a negative phase. + \item If Sigma\_Amax > Amin, then additional parameter lines for each age up to Sigma\_Amax \end{enumerate} \begin{longtable}{p{1cm} p{1cm} p{1cm} p{1.25cm} p{1.25cm} p{1.25cm} p{1.2cm} p{1.2cm} p{1cm} p{1cm} p{1cm}} @@ -2486,7 +2485,7 @@ \subsubsection{Predator Fleet Mortality} \Tstrut & & & & & Sigma & Use & Len(1)/ & & Before & After \\ Fleet & Ymin & Ymax & Amin & Amax & Amax & Rho & Age(2) & Phase & Range & Range \Bstrut\\ \hline - 1 & 1979 & 2015 & 2 & 10 & 1 & 1 & 2 & 5 & 0 & 0 \Tstrut\\ + 1 & 1979 & 2015 & 2 & 10 & 1 & 0 & 2 & 5 & 0 & 0 \Tstrut\\ \hline & & & & PRIOR & PRIOR & & & & & \Tstrut\\ LO & HI & INIT & PRIOR & \gls{sd} & TYPE & PHASE & \multicolumn{4}{l}{LABEL} \Bstrut\\ @@ -2494,23 +2493,19 @@ \subsubsection{Predator Fleet Mortality} 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1, first age} \Tstrut\\ 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1, second age} \Tstrut\\ 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1,... age} \Tstrut\\ - -1 & 1 & 0 & 0 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Rho year fleet 1} \\ - -1 & 1 & 0 & 0 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Rho age fleet 1} \\ \hline \multicolumn{11}{l}{\# Additional fleets (e.g., fleet 2) with 2DAR selectivity} \\ \Tstrut & & & & & Sigma & Use & Len(1)/ & & Before & After \\ Fleet & Ymin & Ymax & Amin & Amax & Amax & Rho & Age(2) & Phase & Range & Range \Bstrut\\ \hline - 2 & 1979 & 2015 & 2 & 10 & 1 & 1 & 2 & 5 & 0 & 0 \Tstrut\\ + 2 & 1979 & 2015 & 2 & 10 & 1 & 0 & 2 & 5 & 0 & 0 \Tstrut\\ \hline & & & & PRIOR & PRIOR & & & & & \Tstrut\\ LO & HI & INIT & PRIOR & \gls{sd} & TYPE & PHASE & \multicolumn{4}{l}{LABEL} \Bstrut\\ \hline - 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1, first age} \Tstrut\\ - 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1, second age} \Tstrut\\ - 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 1,... age} \Tstrut\\ - -1 & 1 & 0 & 0 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Rho year fleet 1} \\ - -1 & 1 & 0 & 0 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Rho age fleet 1} \\ + 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 2, first age} \Tstrut\\ + 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 2, second age} \Tstrut\\ + 0 & 4 & 1 & 1 & 0.1 & 6 & -4 & \multicolumn{4}{l}{\#Sigma selex fleet 2,... age} \Tstrut\\ \hline \multicolumn{11}{l}{\# Terminator line of 11 in length indicates the end of parameter input lines} \\ -9999 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\