PT cycle-1 review remediation: dwelling references, household age semantics, provenance metadata

Task metadata

Latest summary

Built and verified PT cycle-1 remediation bundle pt_population_review_cycle1_d828fc79_seed420987 at /home/synthestat/output/runs/PT/pt_population_review_cycle1_d828fc79_seed420987. The bundle addresses the reviewer’s blocking fixes: occupied dwelling household references are populated, HH_SINGLE_Y/M/E age semantics are explicit and matched by birth dates, provenance metadata marks retrieval/licence/source-universe availability per source, and degenerate seeded person attributes are diagnosed. Created synth-reviewer re-review card t_6b5d1390.

Body

Reviewer verdict for parent task t_bfaf5795 is NEEDS_MODEL_FIX on PT run pt_population_review_cycle1_dd672afa_seed420987.

Bundle path: /home/synthestat/output/runs/PT/pt_population_review_cycle1_dd672afa_seed420987
Reviewer handoff: /home/synthestat/workspace/manager_handoffs/reviewer/2026-05-19_164417Z_PT_population_review_cycle1.md

Blocking fixes requested before next QA cycle:
1. synthetic_dwellings.parquet has household_id null for all 8 dwelling rows while validation_report/diagnostics claim SPT_DWELLING_HOUSEHOLD_REF passed for occupied dwellings. Populate dwelling household_id consistently with households/assignments, or change validation/diagnostics so they do not overclaim this check.
2. HH_SINGLE_E/M/Y household types are emitted, but all persons have birth_date 1991-06-15 (age 34 at 2026-05-19). Add explicit codebook semantics and make generated persons match the labels, or stop using age-coded household labels in this seeded path.
3. source_provenance.json has source IDs/schema hashes but no per-source retrieval timestamps, licence/licence fields, or source-universe fields; live PT INE probe/download was disabled. Strengthen provenance metadata or explicitly mark manual/seeded timestamp/licence/source-universe unavailability.
4. Add/strengthen diagnostics for degenerate seeded person attributes (same birth_date, education, employment, occupation, industry, income bracket for every person) so they are not overread as distribution realism.

Do not treat unavailable hidden/work-school layers as silent failures; they were explicit and acceptable for reviewability, but remain source gaps for broader release.

Runs

Events

Comments

Modeller remediation handoff:
{
  "country": "PT",
  "new_run_id": "pt_population_review_cycle1_d828fc79_seed420987",
  "bundle_path": "/home/synthestat/output/runs/PT/pt_population_review_cycle1_d828fc79_seed420987",
  "builder": "/home/synthestat/workspace/manager_handoffs/modeller/build_pt_review_bundle.py",
  "handoffs": [
    "/home/synthestat/workspace/manager_handoffs/modeller/2026-05-19_1652_build_status.md",
    "/home/synthestat/workspace/manager_handoffs/modeller/2026-05-19_1652_missing_requirements.md",
    "/home/synthestat/workspace/manager_handoffs/modeller/latest.md"
  ],
  "tests_run": [
    "python workspace/manager_handoffs/modeller/build_pt_review_bundle.py",
    "custom Python verification for bundle contract/remediation assertions",
    "python -m pytest tests/test_pt_country_execution.py -q",
    "python -m py_compile workspace/manager_handoffs/modeller/build_pt_review_bundle.py"
  ],
  "remediations": [
    "occupied synthetic_dwellings.household_id backfilled from household dwelling links",
    "HH_SINGLE_Y/M/E age codebook added and synthetic_persons.birth_date aligned to Y=18-34/M=35-64/E=65+",
    "source_provenance.source_records now explicitly mark retrieval timestamp/licence/source-universe availability for each manual source",
    "distribution_diagnostics.synthetic_attribute_degeneracy warns on remaining repeated seeded/modelled person attributes"
  ],
  "review_card": "t_6b5d1390"
}