Edna Free-Living Study: participant-level engagement and dietary-recall timing dataset

This dataset accompanies the MEMDA Free-Living Study, a 14-day study of a mobile dietary-assessment app run under free-living conditions. It contains one row per participant, with app-engagement metrics, dietary-recall timing measures, and demographic characteristics. It is prepared for open reuse under the FAIR principles (Findable, Accessible, Interoperable, Reusable). WHAT IS IN THIS DEPOSIT data/MEMDA_FreeLiving_participant_data.csv — The dataset: 135 participants × 42 variables, one row per participant. UTF-8, comma-delimited. MEMDA_FreeLiving_codebook.csv — Human-readable codebook: every variable with a label, description, data type, units, allowed values, and missing-value code. datapackage.json — Machine-readable schema (Frictionless Data "tabular-data-package"), including field types, primary key, and a SHA-256 checksum of the data file. CITATION.cff — Citation metadata for automated reference managers. STUDY AT A GLANCE Participants used a smartphone app to record their meals over 14 days under one of two prompting modes. In Event mode, the participant opened the app and logged meals themselves. In Interval mode, the app prompted the participant at intervals to report whether they had eaten. The study measured how the two modes affected engagement (how often and how long people used the app), dietary-recall timing (how long after eating a meal was reported), and reporting burden (prompts that did not correspond to eating). Only participants who completed the full 14 days are included in this release. Participants: 135 completers (68 Event, 67 Interval). Age range: 19–65 years. Sex: 71 female, 64 male. App version: 2.1.53.daf1259. VARIABLE SUMMARY The 42 variables fall into four groups; see the codebook for the full definition of each. Identifiers and design: SubjectId, Study, RegisteredMonth, Mode, SoftwareVersion, and the randomisation strata (StrataSex, AgeStratum, BmiStratum, RegionStratum). Engagement and usage: DaysTracked, EatingDaysTracked, Engagement, StudyAvgDailyTimeMinutes, TotalPromptCount, IntervalCompliance, meal and entry counts, and per-day rates. Recall timing: StudyAvgRecallIntervalMinutes, StudyAvgRecallIntervalFinishedMinutes, StudyAvgLatencyMinutes, StudyAvgTimePerMealMinutes, RecallGroup. Demographics: Age, BMI, Sex, Race, Ethnicity, Education, Region, CensusRegion, DemoTimeZone. MISSING VALUES Missing values are coded as NA. Several timing and compliance variables are defined only for a subset of participants (for example, IntervalCompliance and StudyAvgLatencyMinutes apply to Interval-mode participants only), and are NA elsewhere by design. The codebook notes these cases. DATA PROCESSING The released file was derived from raw app exports with a documented pipeline. Data were restricted to study completers (reached day 14). Meal-level timing outliers (3 SD or more within mode) were set to missing before participant-level means were computed. Meals were counted only when at least one food was entered, and consecutive eating meals logged under 15 minutes apart were merged into one. Participant-level summaries were then computed over assigned study days. PRIVACY AND ETHICS The dataset is de-identified. Participants are represented by pseudonymous IDs (MFxxx) that cannot be linked to names, contact details, or device identifiers, none of which are included. Direct identifiers collected during screening (name, email, phone) were never part of this file. The registration timestamp has been generalised to month (RegisteredMonth, YYYY-MM) so it cannot act as a quasi-identifier. Age and BMI are included as continuous values because they are central to the research questions and, combined with the coarsened geography, present low re-identification risk.