Data method
Source acquisition
Every source begins with an acquisition record: origin, date obtained, access terms, owner and whether redistribution is permitted.
Original files kept unchanged
Original files are hashed and preserved in protected raw storage when rights permit. When a source is corrected it becomes a new acquisition that names the version before it.
One common data format
Sources arrive in whatever shape their authors chose. Each is converted into a single common format. The original identifiers are kept, along with a record of every change made, so any converted value can be traced back to the value in the source.
Units and pressure reference
Absolute pressure is stored in pascals, time in seconds, temperature in kelvin and gas content as a decimal fraction. Depth in metres of sea water is calculated from the stored pressure and is provided for readability. The models themselves use pressure. Each dataset records the conversion it used.
Missing data
Missing values stay missing unless a documented derivation is scientifically justified.
How health outcomes are recorded
Several kinds of outcome are kept apart: exposure-only records, self-reported symptoms, medically assessed outcomes, experimental outcomes and observations of venous gas emboli, including Doppler measurements. Each kind stays in its own field and is never combined with the others.
Licensing and access
A dataset can be scientifically useful even when its raw records cannot be republished. Each dataset states metadata visibility, research access and public redistribution rights separately.
Quality control and release
Normalised datasets pass schema, unit, range, duplication and provenance checks before they can be frozen for a research release. Corrections create new versions.