Reproducibility in Practice: Data, Code and What Another Lab Can Rerun
Package data, code, environment information and instructions so another researcher can rerun an analysis. Respect access restrictions, follow repository metadata requirements, and distinguish same-data reproduction from new-data replication.
What a learner can do afterwards
- Package an analysis so a stranger can regenerate a stated figure from the raw data
- Distinguish reproducing an analysis from replicating an experiment, and say what each tests
- Choose a suitable repository and follow its documented metadata requirements for a stated data type
1 · Read
Computational reproduction obtains consistent results from the same input data and workflow. Supply inputs, code, environment, processing, commands and expected-output checks. Install Python 3.11 with its installer. In a terminal, check python --version shows 3.11. From the folder containing requirements.txt, run python -m pip install -r requirements.txt to install its listed dependencies. Identifiable participant data need not be public; document authorised access when restrictions apply.
To check reported values, rerun the workflow. Example: put raw.csv and clean.py version 1.3 in one folder. Restore Python 3.11 and recorded dependencies. If the script reads raw.csv and writes figure.png, run python clean.py there; compare figure.png with expected output. Document paths, processing, relevant seeds and settings. Hardware or numerical libraries can affect output. Pins alone do not guarantee bitwise equality; use a justified tolerance when needed.
Terminology varies. In this lesson, reproduction means rerunning the same-data workflow. Replication means collecting new data to address the same question. Reproduction checks whether the workflow regenerates its reported output. Replication adds new observations under a stated design. Neither guarantees a scientific claim or establishes a researcher's honesty.
A successful rerun can repeat the same flawed assumptions or coding decisions. It checks output under the conditions tested, not every scientific conclusion. Identify code with an immutable version or commit identifier. A moving branch name such as main may later point to different code and cannot by itself identify the version used.
Choose a repository that accepts the data type and supports suitable preservation and access. Raw sequencing reads usually suit a sequence archive; tables and code may suit a general research repository. Follow its published metadata requirements for samples and measurements. Sensitive data may need controlled access. Check consent and applicable restrictions before deposition.
A README should give input paths, environment setup, command order, outputs and comparison checks. A data dictionary defines column meanings, units, scales, category codes and missing-value codes. For example, -999 may mark missing data rather than a measurement. Code can run successfully while a reader misinterprets an undocumented input.
Document inputs, code, environment, commands and checks. Respect access restrictions. Same-data reproduction checks the workflow; new-data replication adds observations.
Original lesson, questions and flashcards about reproducible research workflows, written for this learning dataset.
2 · Watch
به انگلیسیTake it off screen
Where it sits
Optional connections
This opens up
Nothing builds on it yet.
Where this leads
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.