Data quality
The EnaResearch Data quality page - completeness per measurement type and participant, and the implausible, sudden jump, source conflict and low wear flags.
- Admin panel
The Data quality page shows, for each measurement type in the study, how complete the data is and which data looks suspicious. Flags never change values: they are computed when the data is read, using versioned rules, and exports carry them as columns so you can decide in your analysis what to do with flagged data.
Required role
- A member of the study team.
Open the report
Open the study → Data quality and choose the Period: 30 days, 90 days or 1 year.
Measurement types
| Column | Meaning |
|---|---|
| With data | Participants who have data for this type |
| Completeness (median) | Share of days in the period that have data, median across participants |
| Implausible | Samples outside the plausible range (counted per sample) |
| Sudden jump | Samples with a sudden jump from the same source (counted per sample; a single outlier is counted once) |
| Source conflict (days) | Days on which the daily totals of two sources differ clearly |
| Low wear (days) | Days with too few hours of data, for example a watch not worn |
Completeness is counted like the data coverage strip: today and the first and last day of the participation count in the denominator too. Partial days are left out of source conflict and low wear.
Participants
Active, completed and lost-to-follow-up participations with their completeness (the median across measurement types). Participations whose data deletion was requested are left out. Details per participant are in the participant panel.
Quality rules
The Quality rules (version …) section lists the exact thresholds in use, for example "more than … units within … s from the same source". When a rule changes, its version goes up, and each export manifest records the rules it used.
Quality and rejection are different
Values that are physically impossible (outside the catalog's accepted range) are rejected on arrival and never stored. Values that are possible but suspicious are flagged and kept. When you define a variable, you choose whether flagged samples are included; they are left out by default. See Variables.
In exports
Measurement exports carry the flags: a flags column on raw rows, a flagged count on minute, hour and day rows, and a quality.csv file with completeness and flag counts per participant and measurement type. See Exports.
Related articles
Can't find what you need?
Our support team can help, from setting up a study to exporting a dataset. Please include your organization code.
Data flow
How data travels from Apple Health or Health Connect through the EnaResearch app to the server, when it is sent, how it is checked and how to see it arrive.
Manual measurements
Enter pain and fatigue scores, Borg exertion and lab values such as hemoglobin and CRP for EnaResearch participants, and correct or delete them with history.