Skip to main content

The TADA-Wiki: Best practice for sharing analytical code

11 Glossary

Plain-language definitions of the terms used across this wiki. Terms tagged SORTEE follow the glossary in the SORTEE guidelines (Table 1; Pick et al. 2026). The rest are our own definitions. Where a chapter covers a term in depth, the definition links to it.

Absolute path
A file path written out in full from the top of one computer’s file system, such as C:/Users/ed/Desktop/data.csv. Also called a local or hard-coded path. It only exists on that machine, so code that uses one fails for everyone else. Compare relative path. See Transferable.
Analytical code
Code written to analyse one dataset for one study. Its job is to be a transparent, reproducible record of that analysis, not a tool for other people. Compare software. See TADA at a glance.
Annotated
The fourth TADA principle: comments inside the code say what each chunk does, why, and where its output appears in the paper. See Annotated.
Archiving SORTEE
Depositing data and code in a public repository. Archiving is not the same as putting files on GitHub or in supplementary material; see immutable and persistent identifier.
Available
The second TADA principle: the code is publicly archived with a persistent identifier that is cited in the paper. See Available.
CC0 and CC BY
Two Creative Commons licences, common for data. CC0 places work in the public domain with no conditions. CC BY lets others reuse the work as long as they credit you. Neither is designed for code. See Licences.
Code review
Checking and evaluating each other’s code, either formally during peer review or informally among co-authors and colleagues. See Code review.
Computational reproducibility
Obtaining consistent results using the same input data, computational methods and conditions of analysis, the definition the SORTEE guidelines use. In practice, it means getting the paper’s results by running its archived code on its archived data. This is Stage 6 of the SORTEE guidelines.
Container
An image that packages a complete software environment, including all the programs and libraries your code needs, so it runs the same way on another computer. Docker and Apptainer are the common tools. Beyond the scope of TADA.
Copyleft licence
A licence that requires anything built from your code to be released under the same licence. The GPL is the usual example. Compare permissive licence.
Data and code quality control SORTEE
Checking whether data and code are suitable for public archiving. It is not a code review: it does not judge whether the analysis is correct. See The SORTEE guidelines.
Data availability statement
The section of a paper that says where the data and code are archived and how to access them, usually with their DOIs.
Data dictionary
A file listing every variable in a dataset with its description and units. Also called a codebook. A form of metadata.
Data editor SORTEE
An editor at a journal whose job is to screen and quality-control the data and code that will be archived alongside papers under review.
Data filtering SORTEE
Removing some data to create the dataset used in the analysis, for example removing individuals with a certain characteristic. The result is filtered data.
Data processing SORTEE
Transforming data from one form to another: extracting data from images or videos, summarising, transforming or calculating new values. The result is processed data. Typing up a field sheet does not count.
Dependency
A package or library that your code needs in order to run.
DOI
Digital Object Identifier. A persistent identifier for a published object, such as a paper, dataset or code deposit, written like 10.5281/zenodo.1234567. Adding https://doi.org/ in front gives a link that keeps working even if the object’s web address changes. Repositories such as Zenodo and Figshare assign one when you deposit. See Available.
Documented
The third TADA principle: a README and a licence describe the code, its software environment, where the data are and how to run it. See Documented.
Double-blind review
Peer review in which neither the authors nor the reviewers know who the others are. Code and READMEs have to be anonymised for it. See Anonymising for double-blind review.
Embargo
A period during which an archived deposit exists but is not yet public. Many repositories offer one for sensitive data.
Environment manager
A tool that records and restores the exact package versions a project uses, usually in a lockfile. renv (R) and conda, poetry and pipenv (Python) are examples. Beyond the scope of TADA.
FAIR principles SORTEE
Findable, Accessible, Interoperable and Reusable: principles for sharing data and code (Wilkinson et al. 2016). See How TADA maps onto FAIR.
FAIR4RS
The FAIR principles adapted for research software (Chue Hong et al. 2022).
File extension
The ending of a file name, such as .R, .py or .csv, which tells software what kind of file it is. See Transferable.
GitHub
A website for developing code with version control and collaborating on it. It is not an archive, because files can be changed or deleted and it does not give a persistent identifier. See Why GitHub is not an archive.
IDE
Integrated development environment: an application for writing and running code, such as RStudio, Positron, VS Code, PyCharm or Spyder.
Immutable
Cannot be changed or deleted once published. In an immutable repository, such as Zenodo, an update creates a new version instead of overwriting the old one, so the exact files behind a paper stay available. GitHub and OSF are not immutable.
Interoperable
Able to be opened and used across different software and operating systems. Plain-text formats such as .csv for data and .R or .py for code are interoperable; .xls, .docx and .pdf are not. The “I” in FAIR.
Licence
A statement of what others may legally do with your code or data. Without one, others cannot lawfully reuse it, even if they can see it. See The licence.
Lockfile
A file written by an environment manager (for example renv.lock) that records the exact version of every package a project uses.
MCMC
Markov chain Monte Carlo, a family of stochastic sampling methods, common in Bayesian models. Results differ slightly between runs unless a seed is set.
MeRIT
Method Reporting with Initials for Transparency: adding the initials of the person who did each step to the methods section, for example “Code was checked by E.I.C.” (Nakagawa et al. 2023).
Metadata SORTEE
A description of, and information about, the data and code. Usually a text file called a README, sometimes with a data dictionary.
Namespace
The package a function belongs to. Calling a function with its namespace, as in dplyr::filter(), shows the reader which package it comes from.
PEP 8
The standard style guide for Python code.
Permissive licence
A licence with few conditions, usually just that the licence and copyright notice are kept. MIT, BSD and Apache 2.0 are examples. Compare copyleft licence.
Persistent identifier (PID)
A globally unique, long-lasting reference to a digital object that keeps pointing to it even if it moves. DOIs are the most common kind; others include SWHIDs, ARKs, Handles, RRIDs and accession numbers such as those from GenBank. A GitHub URL is not one.
Processed data
Data that have been through data processing.
Project SORTEE
The collection of files archived for one manuscript. GitHub calls this a repository, which causes some confusion; see repository.
Proprietary format
A file format controlled by one company and needing its software to open, such as .xls or SPSS .sav. The opposite of an interoperable format.
Pull request
On GitHub, a proposed change to a repository that others can review and comment on before it is merged.
Raw data SORTEE
Unprocessed and unfiltered data, including raw files such as photos, audio recordings, videos and data sheets.
README
A plain-text file (README.md or README.txt) at the top of a project that says what the project is, what each file does and how to run it. See The README.
Relative path
A file path written from the top of the project folder, such as data/raw/counts.csv. It works on any computer, wherever the folder sits. Compare absolute path.
Release
On GitHub, a named, fixed snapshot of a repository. Zenodo can archive each release and give it a DOI.
Repository SORTEE
A service that stores many projects for the long term, such as Zenodo, Dryad or Figshare. On GitHub the word means a single project instead.
Seed
A number that fixes the starting point of a random number generator, so stochastic code generates the same pseudorandom numbers, and gives the same result, each run. Set with set.seed() in R or random.seed() in Python.
Session info
A printout of the software environment: language version, operating system and every loaded package with its version. Produced by sessionInfo() in R or session_info in Python (from the session-info package).
Software
Code written to be general, modular and tested, so it works across systems and on many datasets. Compare analytical code.
Stochastic
Involving randomness, so the output can differ between runs. Simulations, bootstrapping and MCMC are stochastic.
Supplementary material
Files attached to a journal article. Not an archive: it has no persistent identifier of its own and may not stay open and accessible.
SWHID
SoftWare Hash IDentifier, the persistent identifier used by Software Heritage to reference archived source code.
Transferable
The first TADA principle: anyone can open and run your code on their own computer without converting or editing it. See Transferable.
Unit test
A small automated check that a function gives the expected output for a known input, written with testthat (R) or pytest (Python).
Virtual environment
In Python, a self-contained setup for one project with its own packages, kept separate from the rest of the computer.
Working directory
The folder R or Python treats as its starting point when reading and saving files. Opening an RStudio project sets it to the project folder.
Workflow manager
A tool that runs the steps of an analysis in a defined order and tracks which steps depend on which, such as Snakemake or Nextflow. Beyond the scope of TADA.

Missing a term? See Contribute.