The Transparent Assessment Framework (TAF) is a framework to organize data, methods, and results used in ICES assessments, so they are easy to find and rerun later with new data.
It will provide an online archive of final assessment for each year, for all stock categories. All data input and outputs are linked to existing or upcoming ICES data services.
| URL | Purpose |
|---|---|
| taf.ices.dk | Main landing page |
| doc | Documentation |
| wiki | Wiki |
| ices-taf | Assessments and other analyses |
| diagrams | Introductory diagrams |
| tutorial | Tutorial: Simple assessment |
| github_guidelines | ICES GitHub |
| Package | Links |
|---|---|
| icesTAF | CRAN homepage and help pages |
| TAF | CRAN homepage and help pages |
| ICES pkgs | GitHub repositories and release history |
The TAF stock assessment workflow is scripted in a series of R scripts.
| Core scripts | Purpose |
|---|---|
data.R |
preprocess data, write TAF data tables |
model.R |
run analysis, write model results |
output.R |
extract results of interest, write TAF output tables |
Optional scripts include report.R (prepare plots and tables for the report)
and utilities.R (custom functions used in the above scripts).
If a script becomes long, it is recommended to split it into separate scripts
with the same filename prefix, e.g. report_plots.R and report_tables.R,
where a main report.R script can simply source the underlying scripts:
source("report_plots.R")
source("report_tables.R")The TAF data presentation describes the flow of data
from boot/initial/data to boot/data and finally to data.
Tonnes, thousands of individuals
The boot/initial folder contains the initial data and software that are
not available online from another repository. These files are the basis of all
subsequent analysis, and guarantee that the TAF analysis can be rerun later and
will reproduce the original results, even if data in the underlying databases
(outside of TAF) may have changed after the original analysis was submitted.
One objective of TAF is to document the data preparation. Therefore, the initial data files should represent the original data before the main preprocessing and data aggregation takes place. For example, the initial data could be the result of an SQL database query, selecting one species of interest and all relevant data columns for the analysis.
When using software that is not available online from another repository,
boot/initial/software contains the model source and executable. A Windows
executable is required, while Linux and macOS executables are optional.
See taf.boot and process.bib in: