README.md

In CCSBT/sbtr: sbtr

sbtr

The sbtr repository is an R package for manipulating and plotting Southern bluefin tuna (SBT) model outputs.

Installing sbtr

Installation of the sbtr package can be done from within R using:

# Install package
install.packages("devtools")
devtools::install_github("CCSBT/sbtr")

# Load package
library(sbtr)

Alternatively, the package can be cloned to your computer, then built and installed using:

git clone https://github.com/CCSBT/sbtr
cd sbtr
make

Using sbtr

Once installed, the sbtr package can be called in R using library(sbtr).

All figures produced during the 23-26 July 2013 meeting in Portland, Maine, are contained in the file 2013_07 Portland meeting scripts.r. Typically how the code is organized is with a directory containing all the .r code, and subdirectories \figs where the .pdf figures are saved, \levfiles containing the .lev files, \gridfiles containing the .grid files, \tables where the .csv outputs are saves. One final directory, the \arc directory, contains a subdirectory for each set of model evaluations, including the all-important lab.rep files, e.g. \arc\base2010sqrtns\base2010sqrtns_h1m1M1O2C2a1_lab.rep.

The code works by first reading in all of the lab.rep files for a particular model run (all lab.rep files in a directory) and saving these in a list, e.g. data.base2010ns. This takes a few minutes. All subsequent code uses these lists and runs quickly (1-5 seconds).

There is a .zip file containing all lab.rep and associated files, and a .zip file containing just the .r files.

The word file in the R directory provides further details.

Examples

Example input files are provided in arc/example_lab.rep and levfiles/example.lev. The _lab.rep input file can be read into R using:

data.example <- get.all.files("arc/example_lab.rep")

Reading _lab.rep files takes a long time so it is recommended to save the data object as an .RData file or similar once read in. Now tables and plots can be produced as desired. For example:

nll.table.example <- likelihood.table2(data.objects = data.example)
write.csv(file = "tables/exampleNLL.csv", nll.table.example)

pdf("figs/Fig 1 NLLs example.pdf")
Plot.NLL.by.steepness(nll.table.example, caption = "example")
dev.off()

Alternatively, one can plot .lev files using:

jpeg(file = "figs/Fig 5 Shaded base.jpg", width = 600, height = 600)
Plot.lev("levfiles/example.lev")
dev.off()

Adding R functions to the sbtr package

The sbtr source code is located in the directory sbtr/R/. Additional R functions can be added to the package by placing them in the R/ directory. Generally a single R function should be included in a file with the naming convension function_name.R. For a function to be included in the package a roxygen header must be added above each new R function. This header and function should be formatted as follows:

#' Function title
#' 
#' Description of function
#' 
#' @param y the function input
#' @return something done on y
#' @author John Doe
#' @examples
#' x <- function_name(y) # An example of how to run the function
#' @export
#'
function_name <- function(y)
{
    out <- do_something_to(y)
    return(out)
}

The package can be compiled using the Makefile by simply typing make. This builds the .Rd helpfiles and the NAMESPACE file within the package source directories and then builds and installs the package.

Script for covariance function

To evaluate the within-cell variabilty in addition to the structural accross-the-grid uncertainty.

Preliminaries:

May need to edit main.tpl so that -est in call to sbtmod is removed This needs to be added as a flag rather than hardwired as computing the Hessian each run adds time to evaluating the grid

E.g., within directory OM run: main base2013 sqrt

to create the base2013sqrt subdirectory within OM/arc and save labrep files AND now the .cor .std and .par files that can be used for further evaluation

Below are the scripts that evaluate these results further

install.packages("PBSmodelling")
library(PBSmodelling)
library(sbtr)

# Edit this to your needs...
setwd("/Users/jim/_mymods/sbtmod/runs")

# These two lines take time...1st reads in all results, second reads in and computes covariances
labrep <- get.all.files("arc/base2013sqrt")
covrep <- get.cov.est()

# If your data had been stored in R:            
load("My.RData")

# Use covariance to generate B10 and F simulations of posterior over each point in grid. Also includes terminal year numbers at age estimates and covariances.
sims  <- sim.F.B10(covrep,lev.file="../OM/arc/grids/base2013sqrt.lev")
lev.file="../OM/arc/grids/base2013sqrt.lev"

# Plot comparisons of within cell variability vs grid
MSYcov.stats  <- Plot.MSY.cov(labrep,lev.file,sims,xlim=c(0,2.3))

CCSBT/sbtr documentation built on Oct. 25, 2020, 9:11 p.m.