Changelog
Source:NEWS.md
SAMtool 1.7.0
 Add argument to adjust annual recruitment deviation bias correction in
RCM
in argumentpbc_recdev
. The bias correction from logspace to normal space isexp(log_rec_dev[y]  0.5 * bc_recdev[y] * sigmaR^2)
if the yearspecific rec dev is estimated.  Add
RCMdata@I_delta
slot to specify survey timing (01) within year. Default assumption is zero is used which is consistent with previous versions.
SAMtool 1.6.5
 Update
RCM2MOM
(report selectivity at length array, catch fractions).  Update RCM report to identify nonconverged simulations in figures and data predictions for which there is no likelihood.
SAMtool 1.6.4
CRAN release: 20240215
 RCM passes maturity at length parameters to operating model. While RCM only uses the corresponding age ogive, the OM later will backcalculate to retrieve the length parameters, but may not recover the original values.
 RCM passes selectivity at length to operating model. This is needed to simulate length data in the projection.
 Plot maturity at length and lengthweight relationship in RCM output if fitted to sized data.
 Histograms exclude extreme outliers (samples greater than the 95th percentile if the skewness > 3).
SAMtool 1.6.3 (20231118)
CRAN release: 20231118
 Under
RCM_output
tab inplot(RCModel)
, report simulations and convergence rate, and clean up comp plots.  Fix issue with F calcs when
RCM(condition = "catch2")
, introduced in version 1.6.0.
SAMtool 1.6.2 (20231026)
CRAN release: 20231026
 Minor
RCM
updates to calculate fishery length comp whenCAL_n > 0
and protectplot.RCM
from empty index vectors.  Update roxygen doc with markdown.
SAMtool 1.6.1 (20230823)
CRAN release: 20230823
 Patch updates to tidy up RCM report to prevent empty length selectivity figures.

Perfect
uses spawn_timing to calculate spawning biomass (exp(spawn_time * M)
) in the middle of projection year. Note that perfect HCR implementation needs to iteratively recalculate the projection year biomass, B/BMSY, B/B0 (exp(spawn_time * [Ftarget + M])
) when applying the HCR. Recommend spawn_time = 0 to implement HCR perfectly.  Update URLs to describe how to map index q in RCM.
 RCM no longer uses
ObsPars$Isd
.  Update yield curve calcs for RCM with fecundity and spawn timing.
 Update retrospective function for RCM (map dimensions changed in TMB).
SAMtool 1.6.0 (20230719)
CRAN release: 20230719
RCM updates
 Add
spawn_time_frac
argument to RCM.  Add fleetspecific option for catch/effort conditioning to RCM.
 Allow map argument for index q
RCM(map = list(q = c(1, 1)))
. This example allows sharing q between 2 indices. Currently, q can only be an explicit estimated parameter (map argument is an integer), solved analytically (map argument is NA), or fixed to 1 (map argument is NA and additional specification in RCMdata@abs_I).  Add explicit selectivity for length (“logistic_length”, “dome_length”) or agebased (“logistic_age”, “dome_age”) functions. For backwards compatibility, “logistic” and “dome” map to their lengthbased equivalents.
 Update prior for the length or age of full selectivity to be
x^0.01 * (1  x)^0.01
where x is the ratio of the length of full selectivity to Linf or age of full selectivity to maxage.  Update figures for RCM markdown reports.
SAMtool 1.5.3 (20230523)
CRAN release: 20230523
 Fix indexing issue when estimating steepness posthoc in
SCA2
andVPA
.  Return warning when arguments to
make_MP
don’t match formal arguments in.Assess
and.HCR
SAMtool 1.5.2 (20230404)
CRAN release: 20230404
 Custom internal
hist
function will report NA rate (percent of NA’s) in a vector. Seen in markdown reports.  Internal
max
function excludes infinite values. Primarily used when generating axes limits in markdown reports.  Fix bugs when
Data
object is passed toRCM
.  Allow
resample = TRUE
with stochastic fits toRCM
.
SAMtool 1.5.1 (20230208)
CRAN release: 20230208
 Add gridlines to markdown plots for easier viewing. Update some plots and reorganize tabs in RCM markdown report.
 Add a Gaussian penalty of
dnorm(log(Shinge/min(SSB)), 0, 2)
for hockeystick SRR when the hinge point is less than the smallest SSB.  Convert zeros to 1e8 when found in
RCMdata@Chist
. The trivially small catch still allows predictions of fishery age composition from Baranov equation.  Turn various RCM messages to warnings, e.g., for nonconvergence, high F, etc.
 Fix bug in
diagnostic
introduced in 1.5.0.
SAMtool 1.5.0 (20230129)
CRAN release: 20230129
 Add
simulate
method for RCM and assessment models.  New
map
andstart
arguments forRCM
.  Add progress bar with timers for RCM and various diagnostic functions using
pbapply
.  Add hockeystick stockrecruit relationship in
RCM
using the Mesnil and Rochet (2010) parameterization.  Users can now add lognormal priors for
r
andMSY
for surplus production models.  Update reporting of assessment output inside MSE. Use of data frames instead of lists should save space in the
MSE
object.
SAMtool 1.4.1 (20221116)
CRAN release: 20221116
 Fix
Data@CAL
check when usingRCM
.  Use
Gmisc::fastDoCall
when fitting models, e.g.,SP_Fox
.Gmisc
is aSuggests
package.
SAMtool 1.4.0 (20220607)
CRAN release: 20220607
 RCM can now use multivariate logistic and Dirichlet multinomial distributions for fitting to age/length comps.
 Compile with TMBAD library.
 Import individual functions in NAMESPACE to avoid package conflicts.
SAMtool 1.3.1 (20220506)
CRAN release: 20220506
 Patch in RCM markdown report for plotting fishery length composition.
SAMtool 1.3.0 (20220408)
CRAN release: 20220408
 Require TMB 1.8 (need TMB package consistency between compilation and model fitting)
SAMtool 1.2.6 (20220331)
CRAN release: 20220331
New features

RCM2MOM
converts the output ofRCM
to a multifleet operating model.  Vignette is now hosted online.
 Added a beta version of
RCM_assess
for using the RCM model as an assessment in closedloop projections. More arguments will be added in the future for flexibility with model configuration. 
make_project_MP
creates management procedures that update TAC annually from stock assessment projections. 
posterior
wrapper function added to run MCMC of RCM models.RCMstan
updates OMs with MCMC output.
Minor updates
 Fix FSPR reporting
Shortcut
andPerfect
assessment functions.  Fix F calculation in
HCR_segment
with yield per recruit (F01 and Fmax).  Updates to
interim_MP
include adding NULL catch for catch advice and adding missing feature to report assessment output whendiagnostic = 'full'
.  Create ability to specify output F in absolute magnitude (independent of model output) in
HCR_segment
andHCR_ramp
.  Uniform prior available for
R0
andlog(R0)
for RCM models and assessment models.  Shared recruitment deviation parameters in
RCM
only enter the objective function once.  New correlation plots of estimated productivity parameters (steepness, M, R0, as well as depletion) in
RCM
reporting.  Clean up TMB code to fix clang14 compiler warnings.
SAMtool 1.2.5 (20220112)
CRAN release: 20220112

SCA_RWM
can accept multiple years to therefyear
argument, e.g.,expression(1:Data@Year)
. The model will calculate reference points (MSY, unfished values, and steepness) using the mean M during the specified years.  Fix bug in interim MPs to return
NA
inRec@TAC
when multiple assessments do not converge.
SAMtool 1.2.4 (20211215)
CRAN release: 20211214
 Update
Shortcut
indexing to align year of assessment with projection. An MP using thePerfect
assessment andHCR_MSY
annually will produce F = FMSY in the OM.  Stabilize
VPA
when the catchatage in the plusgroup and plusgroup1 is very small.
SAMtool 1.2.3 (20211128)
CRAN release: 20211128

RCM
will check age and length comp data for NA’s and replaces with zero 
RCM
reports annual equilibrium unfished reference points using constant stock recruit alpha and beta  Unlink TMB executable when unloading package
 Fix punctuation issue in markdown reports that use knitr v1.36.
 Other minor fixes
SAMtool 1.2.2 (20210916)
CRAN release: 20210916
 The
make_interim_MP
function is added to generate MPs that adjust the TAC between periodic assessments using the index.  An additional posfun to
SP
is added to avoid negative biomass situations.  Fix sampling of recruitment deviations for projections in
RCM
so that the mean is one in normal space. This error was apparent when autocorrelation was very large.  Assessment and RCModel objects now save the package version as attributes.
 Shortcut function now reports yearspecific reference points for the harvest control rule.
 HCR returns TAC = NA when a control point or target point is negative.
SAMtool 1.2.1 (20210812)
CRAN release: 20210812

HCR_segment
allows for creating control rules with any number of linear segments.  Assign index beta = 1 for OMs conditioned by RCM.
 The ESS argument is reinstated for RCM to maintain backwards compatibility.
 Fix assignment of length bins in OM by
RCM
.
SAMtool 1.2.0 (20210706)
CRAN release: 20210706
RCM
 A new S4 object,
RCMdata
, is used to send data to the RCM model, i.e.,RCM(OM, RCMdata)
. For now, backwards compatibility should still be maintained when feeding a data list (used prior to v1.2) to fit the model.  Internal R and TMB code for RCM has been revised, e.g., reducing interchangeability between the terms ‘survey’ and ‘index’ to focus on ‘index’ as much as possible when maintaining backwards compatibility.
 The
profile
generic is now available forRCM
models. Steepness, R0, and final depletion can be profiled.  Fixed plotting bug in
compare_RCM
.  Priors for index q in
RCM
are now lognormal instead of normal.  Annual equilibrium and transitional SPR are now calculated and reported.
 Uneven length bin widths are now supported.
 An example is added to the help documentation and online article, using Pacific cod (courtesy of R. Forrest).
Assessment models
 Priors on M, steepness, R0, and index q (lognormal, see RCM) can be created SCA, DD, and cDD assessment models.
 Likelihood weights to SCA assessments can now be provided for
Catch
,CAA
, andCAL
in addition toIndex
in a named listLWT
. Backwards compatibility remains to provideLWT
as a vector for index likelihood weights only.  A new SCA assessment model that incorporates densitydependent natural mortality (
SCA_DDM
) is added.  A new SCA assessment model that fits to length composition (
SCA_CAL
) is added.  Delaydifference assessments (DD_TMB, DD_SS, cDD, cDD_SS) can now fit to mean weight with argument
MW = TRUE
. The functions will look for mean weight data series inData@Misc[[x]]$MW
, otherwise will convert length compositionData@CAL
to weights and calculate annual means.  Delaydifference assessments (DD_TMB and DD_SS) now use an instantaneous F formulation for the catch equation instead of Pope’s approximation. The models should now be more robust for high F situations.
 Dynamic SSB0 is calculated for all assessment models.
 A variant of the shortcut assessment emulator for closedloop simulation is available (
Shortcut2
). This function fits an SCA assessment and then characterizes the assessment error relative to the operating model using a vector autoregressive (VAR) model. The functions samples the operating model with error predicted from the VAR model for the projection period. This is a useful function to guide the level of error in the shortcut method.
SAMtool 1.1.2 (20210309)
CRAN release: 20210309
 Minor fix to vignette to fix MSEtool reverse dependency issue.
SAMtool 1.1.1 (20210217)
CRAN release: 20210217
 Fix year range for depletion calculation from timevarying SSB0 in RCM.
SAMtool 1.1.0 (20210202)
CRAN release: 20210202
 Edits to fix valgrind and sanitizer issues in TMB code.
 Likelihood gradients, the derivatives of the likelihood of each annual data point with respect to model parameters, are plotted in the RCM markdown report. For this purpose, the annual age or length composition is considered to be a single piece of data. This diagnostic could be informative on how informative the data are to model parameters, with more influential data points having larger gradients.
 When conditioned on effort, the
RCM
will now incorporate catches into the likelihood as a default. This allows the model to estimate F and R0 when conditioned on effort and there is patchy catch data.
SAMtool 1.0.0 (20210122)
CRAN release: 20210122
 Development of the assessment models and OM conditioning model in SAMtool 1.0.0 continues from MSEtool 2.0.1.
multiMSE
remains in MSEtool.  The age structure of the SCA models (
SCA
,SCA_Pope
,SSS
) start at age 0 following the change in the MSEtool OM.  An additional SCA model (
SCA_RWM
) can be used to estimate timevarying M (constant with age) as a random walk. Fix the random walk SD to a low value to effectively estimate a timeconstant M (see help page).  Warnings during the fit of the assessment models (through
nlminb
) are turned off. Convergence status and issues can be checked in theconv
slot of the output Assessment object. In closedloop simulation, thediagnostic
function can be used to track the behavior of modelbased MPs. By default, prepackaged modelbased MPs and MPs made from themake_MP
function are designed to report convergence info (stored inMSE@PPD
).  Assessment functions now calculate and report spawning potential ratio and yield per recruit in the forecast slot of the S4 object. Also in this slot is a catch equation function calculates the TAC for a given F.
 HCR nomenclature has changed. Operational control points (OCPs) are used instead of reference points (to help distinguish between reference points in the estimation model vs. the operating model. Various types of F can now be used in the HCR, including F0.1, Fmax, and FSPR, in addition to FMSY for the TAC calculation.
 A
Shortcut
assess function samples the OM with error and autocorrelation for HCRs as an emulator of a stock assessment in closedloop simulation. ThePerfect
function samples the OM without error.  All assessment models now accommodate multiple indices in model fitting, specify in the
AddInd
argument of functions which index slots in the Data object will be used among Data@Ind, Data@SpInd, Data@VInd, and Data@AddInd. Within series weighting is applied by using the corresponding CV slot, i.e., Data@CV_Ind for Data@Ind, etc. Among series weighting can also be tuned using likelihood weights withLWT
argument. For SCA and VPA models, the selectivity is fixed in the model using Data@AddIndV for indices in Data@AddInd.
RCM
 The function for the OM conditioning model is now renamed to
RCM
(Rapid Conditioning Model).  The age structure of the model now starts at age 0, following the change in the MSEtool OM. The dimension associated with age in matrices and arrays need to be of length
maxage + 1
which corresponds to ages 0 to maxage.  When estimating fleet F in the model (
condition = "catch"
), the likelihood for the catch can now have a userdefined standard deviation indicated indata$C_sd
(year and fleet specific, the previous default was 0.01 was builtin for all catches).  For generating length comps, variability in lengthatage can now be agespecific. Specify the lengthatage standard deviation in
OM@cpars$LatASD
.  Priors for log_R0 (normal distribution), steepness (beta distribution for BevertonHolt, normal for Ricker), log_M (age and time constant, normal), and survey q (normal) can now be specified.