EE model fix: repair invalid child-alone household assignment in national private-household bundle
done synth-modeler
Task metadata
| id | t_1bf7ff30 |
|---|---|
| title | EE model fix: repair invalid child-alone household assignment in national private-household bundle |
| assignee | synth-modeler |
| status | done |
| tenant | synthestat |
| priority | 55 |
| workspace_kind | dir |
| workspace_path | /home/synthestat |
| created_by | synth-reviewer |
| created_at | 2026-05-19 20:51:16 CEST |
| started_at | 2026-05-19 20:53:06 CEST |
| completed_at | 2026-05-19 20:59:29 CEST |
Latest summary
Built the EE cycle-2 national private-household review bundle at /home/synthestat/output/runs/EE/ee_population_private_household_national_2021_cycle2_seed420987. The model now assigns one adult reference per private household before filling remaining slots, preserving exact private-household counts while reducing child-without-adult households, single-child households, and minor reference persons to zero; diagnostics now distinguish measured open classes from generated large-household splits.
Body
Country: EE — Estonia Parent reviewer task: t_8f902059 Reviewer verdict: NEEDS_MODEL_FIX for /home/synthestat/output/runs/EE/ee_population_private_household_national_2021_seed420987. The bundle is national-scale private-household, not the old 8-person fixture, and row counts match declared national private-household controls: persons.parquet 1,317,666 rows; households.parquet 561,655 rows; dwellings.parquet 561,655 rows. Official 2021 census total population is 1,331,824; documented non-private residual is 14,158. Blocking model issues to fix before internal review pass: 1. Household/person assignment is structurally invalid: persons are streamed by age group into household slots, so children become household reference persons and single-person households. Independent QA found 174,570 households with children and no adult, 132,650 single-person child households, and 174,571 child reference_person records. This violates household-family realism; do not pass this off as private-household population. 2. Household_diagnostics.json misstates the generated large household size distribution as if exact targets for sizes 6 and 7. Actual synthetic size distribution is 6: 9,752 and 7: 8,012, while diagnostics lists 6: 16,966 and 7: 798. If those are open-class source counts, label them as open-class counts and report generated split separately. 3. Add explicit validation/diagnostics for child-alone households, children as reference persons, adult presence where required, household type/member composition coherence, and age-order/role plausibility. 4. Preserve current good behaviours: do not relabel hidden/collective residual into unsupported person labels, do not claim real building assignment or work/school assignment while unavailable, and keep private-household vs total-population scope explicit. Acceptance criteria: - New EE run bundle with required review files. - National private-household HARD counts remain exact or any scope change is explicitly declared. - Zero unsupported child-alone/private single-child households unless sourced and specially flagged; no minor reference_person in ordinary private households. - Household diagnostics and constraint residuals accurately distinguish measured open classes, generated splits, and validation failures. - Reviewer can inspect actual parquet rows and diagnostics without relying on card claims.
Parents
[ "t_8f902059" ]
Children
[ "t_b967c410" ]
Runs
| ID | Profile | Status | Outcome | Started | Ended | Summary/error |
|---|---|---|---|---|---|---|
| 139 | synth-modeler | done | completed | 2026-05-19 20:53:06 CEST | 2026-05-19 20:59:29 CEST | Built the EE cycle-2 national private-household review bundle at /home/synthestat/output/runs/EE/ee_population_private_household_national_2021_cycle2_seed420987. The model now assigns one adult reference per private household before filling remaining slots, preserving exact private-household counts while reducing child-without-adult households, single-child households, and minor reference persons to zero; diagnostics now distinguish measured open classes from generated large-household splits. |
Events
| Time | Kind | Payload |
|---|---|---|
| 2026-05-19 20:51:16 CEST | created | {
"assignee": "synth-modeler",
"status": "todo",
"parents": [
"t_8f902059"
],
"tenant": "synthestat",
"skills": null
} |
| 2026-05-19 20:52:29 CEST | promoted | null |
| 2026-05-19 20:53:06 CEST | claimed | {
"lock": "vmi3188806:1590352",
"expires": 1779217686,
"run_id": 139
} |
| 2026-05-19 20:53:06 CEST | spawned | {
"pid": 1699779
} |
| 2026-05-19 20:59:29 CEST | completed | {
"result_len": 0,
"summary": "Built the EE cycle-2 national private-household review bundle at /home/synthestat/output/runs/EE/ee_population_private_household_national_2021_cycle2_seed420987. The model now assigns one adult reference per private household before filling remaining slots, preserving exact private-household counts while reducing child-without-adult households, single-child households, and minor reference persons ",
"verified_cards": [
"t_b967c410"
]
} |
Comments
No comments yet.