Exports and data

The province's data goes out as plain tables, each with a data dictionary and its provenance: the app and its version, when it was made, the simulated span, the seed, the basis and assumptions hashes, and every parameter that differs from the defaults. A result can always be traced back to what produced it. The Exports chip in the top bar (or View › Exports) opens them.

Exports: the province's mortality experience, the person-year panel, the burial society's model points and the economy, each to CSV, JSON or a folder of the workspace.

What you can export

Export	Table(s)	One row per
Mortality experience	unity_mortality_experience: year, age, age_width (banded), sex, the group (if any), person_years, deaths, expected_deaths, qx_basis	Calendar year × age (single, or five-year bands) × sex, optionally by city, settlement or tier
Who dies: the person-year panel	unity_person_years: person_id, year, age, sex, person_years, death_event, and the person's city, settlement, tier, heritage, birthplace, schooling, job, income, household size, poverty, cover, chronic conditions, HIV status and vitality; and unity_deaths, the deaths register (cause included)	Resident × calendar year; death
The burial society's book	unity_burial_society_model_points: policy_id, household_id, cover (funeral or life), age_at_entry, sex, sum_assured, policy_term, duration_mth, premium_pp	Covered life × cover
The economy, month by month	unity_economy_monthly: CPI and inflation, the repo and prime rates, nominal and real GDP, unemployment and employment, the Gini of income and wealth, tax revenue, government spending, the fiscal balance, household disposable income, the saving rate, bank deposits and loans, the petrol price, Brent and the rand	Month

The experience and person-year exports carry the true basis the province lived on (the table by single age and sex, its improvement, and the individual risk factors), so a table fitted to the data can be scored against the answer. The policy and stress lab exports its experiments the same way.

Formats

Every export offers three:

Format	What you get
CSV	The first table, as RFC 4180 CSV (a header row; empty cells for missing values). For the person-year export that is the panel, without the deaths register.
JSON	The whole export in one file: every table, the provenance and the true basis.
Save to workspace	With a workspace open: a folder exports/<name>-<date>/ holding every table as CSV, a README.md (the provenance, the headline figures, the true basis and a data dictionary for every table), export.json, and truth.json when the export carries a true basis.

Experience, measured properly

Exposure in the experience exports is central exposure, rebuilt person by person: a resident counts on every day from their first in the province (the start, their birth or their arrival) to the day before their death or departure, at their age last birthday that day, cut at each 1 January. Expected deaths are the basis's daily hazard summed over the same days, with improvement and without individual risk; qx_basis is the basis q, exposure-weighted over the cell. Calendar years observed for less than half their days are left out, because a period model reads every year as a whole one.

One province is about four hundred people, a handful of deaths a year, so most cells of one province's experience are empty. Pool on the worker pool lives the province's basis on many seeds (16 by default, up to 64; 10 years by default, up to 40; at most 800 province-years) and pools their experience into one export. The person-year panel describes each person as at the end of observation (or their death or departure): the engine keeps no history of a job or an income.

Living on a table of your own

Import a mortality table (CSV), at the bottom of the tab, reads a table (see Mortality tables for the format) and rebuilds the province on it, from day 0 on the same seed, keeping its other parameters and any shocks. A/E is then measured against your table, the Exports chip reads Exports · basis, and the card compares your table with the preset at key ages.

Return to the preset basis and lift the shocks goes back: it removes the table and every shock, keeps the other parameters, and rebuilds from day 0.

The format

Every export is a scelo.exchange/1 document: an open, self-describing layout that R, Python, a spreadsheet or Scelo IDE can read as it is. The exchange format describes it.