How to Organize Research Data
Set up a research project folder structure and download a README and data dictionary template with worked examples.
A research project folder structure
Give raw inputs, analysis code and generated results separate folders. Put a README at the project root and a data dictionary next to the data. That way, a colleague can identify the original file and know which outputs the analysis will regenerate.
study-sessions/
├── README.md
├── data/
│ ├── sessions.csv
│ └── data-dictionary.md
├── analysis/
│ └── analyze.py
└── results/
└── means.csv
Keep the original input intact. Put a cleaned or derived file in a clearly named location and document the script that creates it. In this example, sessions.csv is the source and means.csv is generated. A new run can replace the latter safely.
Download the four-row sample dataset. The records are synthetic.
How to write a data dictionary
A column called minutes is a start. Tell readers what it measures, which values are valid, and what to do when a value is missing.
| Column | Meaning | Expected values |
|---|---|---|
session_id | Identifier for one study session | Unique text such as S01 |
condition | Sound setting during the session | quiet or background-noise |
minutes | Recorded session duration | Non-negative whole minutes |
All three columns are required in this four-row example. For real data that can be incomplete, document the missing-value representation and handle it explicitly in analysis. A zero-minute session and an unrecorded session mean different things.
What to put in a research README
Start with the research question and where the records came from. Then list the input, how to run the analysis, and where to check the result. Link to the data dictionary for field-level detail.
This example compares mean study duration under two sound settings using four synthetic records. Open
data/data-dictionary.mdfor column definitions. The input isdata/sessions.csv. Runpython analysis/analyze.pyand openresults/means.csv.
For real data, add collection dates, access instructions and processing notes. Before sharing a folder, check whether it contains identifiers or other restricted records.
Start with the research README template and data dictionary template. Both are Markdown files you can copy into your own project. See how they fit into a runnable reproducibility package.
Practice on your own folder
Find one ambiguous file name in a project, such as clean-final2.csv. Give it a name that describes what it contains, then write down its source and how it was generated. Update any script that reads it.
Check your documentation
Give a colleague the README. Ask them to identify the raw file, the unit used in one column, and the output that can be regenerated. If they have to guess, add the missing detail where they first looked.
References
Adapted from a chapter by The Turing Way Community (CC BY 4.0). Miss Formula reorganized and expanded the explanations and wrote the worked example and exercises.