Simulation design
Where and what, by time of day
Mean hours per day
Diary viewer
Export
Share of the pool by time of day
Mean hours/day — microenvironment
Mean hours/day — activity
Diaries by age
Data
Diaries come from the uploaded Consolidated Human Activity Database (CHAD) extracts: a questionnaire file (one row per person-day, with demographics, date and weather) and an events file (2.2 million activity/location episodes). Every diary covers a full 1,440-minute day starting at midnight. Consecutive events with the same activity and location codes are merged into single episodes (1.01 million).
Donor resampling
Mirrors how SHEDS and APEX assign CHAD diaries. Each simulated person is given an age and gender (drawn from the pool or set by you). For each calendar day, the dashboard picks a random diary from the pool stratum matching that person's age band, gender, day type and season. If the stratum is smaller than the minimum size, matching relaxes step by step and the relaxation level is recorded in the output (match_level: 0 = full match). With the repeat probability r, a day reuses one of that person's earlier diaries of the same day type. With longitudinal donors on, new draws prefer unused diaries from the same CHAD respondent, keeping real within-person behavior where multi-day records exist.
Markov synthesis
Generates new sequences instead of copying real ones. For each stratum, the day is divided into time steps; each step's state is the joint (microenvironment, activity) category at the step midpoint. The model estimates the starting-state distribution and, for every time step, transition probabilities between states. Sequences are sampled from that chain; each run of a state receives specific CHAD activity and location codes drawn in proportion to their time in that state in the stratum. Unseen transitions fall back to the next step's marginal distribution. First-order chains reproduce time-of-day patterns well but tend to make episodes shorter and more fragmented than real diaries — compare the Gantt view against donor results.
Caveats
- CHAD studies differ in design, era (1982–2010), region and diary resolution (e.g., Baltimore records hourly). Use the study and year filters to control what enters the pool.
- Location code 30010 ("your residence", room and indoor/outdoor unspecified) is mapped to Home indoor, as most exposure models do.
- The pool counts are diaries, not people: respondents with many diaries weigh more. Use the repeat/longitudinal settings with this in mind.
- MET values, gas-stove, smoking and ventilation flags are in the source events file but not used here.