Provide the dataset with enough documentation to explain what each row, variable, and code represents. Preserve the original data and record the changes made to the working copy.
Describe the row structure
State whether one row represents a participant, a visit, an admission, a sample, or another unit. For repeated measurements, document how records connect to the same participant and how measurement time is recorded.
Provide a data dictionary
Record the variable name, definition, unit, permitted values, and treatment of missing or inapplicable values. Explain coded categories such as treatment groups and outcome levels.
Organise the handover files
- The original dataset, retained as the source record.
- A working dataset with any cleaning changes documented.
- The study protocol and relevant collection forms.
- The variable definitions and coding scheme.
- Notes about repeated records, exclusions, missing data, and corrections.
- The research question and the analyses required.
Keep data and display formatting separate
Store values in the dataset and put definitions or explanatory notes in the dictionary. Avoid using cell colours or visual layout as the only record of a category or decision.
Original documentation
For software instructions, use the R Project and the Quarto documentation. The structure required for an analysis is reviewed in the context of the study.
Data and statistics services · All learning resources
THOR working guide · Last reviewed: 3 October 2026