From collected items to analytical variables
An assessment item is an actual questionnaire response, laboratory result or measurement captured by a study. A target variable is the concept an analysis needs. Several items may be needed to produce one variable, and participating studies may have used different items to capture the same concept. A DataSchema organizes target variables and their supporting documentation.
For example, one cohort could record height in centimeters and another in meters. Both values can be expressed in a common unit if the definitions and measurement procedures are compatible. Self-reported height and a measured clinical value pose a different question. A conversion can standardize units, while differences in how measurements were collected may require a separate quality decision.
Four review steps
- Specify the target variable, unit, allowed categories and relevant time period.
- Review the instruments and metadata each study collected, recording any limits on comparability.
- Document transformations, recoding rules and missing-data decisions before applying them.
- Run the transformations, check the derived values and retain enough documentation to reproduce the decisions.
Historical documentation
DataSchema structure
Modules, themes and domains create a navigable hierarchy. At the variable level, definitions and reference instruments give reviewers the detail they need to compare data. See the archived DataSchema structure protocol.
Variable selection
Selection starts by defining the purpose of the schema. Subject experts then review relevance, feasibility and the availability of suitable assessment methods before refining the set. See the archived variable selection protocol.
For later guidance, read the Maelstrom Research harmonization guidelines. The historical definitions page records the terminology originally used by the project.