Jinovo/language-conditioned-psychometric-harmonization-model
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

review(stage1): record signed primary outcome reviews

Browse Source 
Both reviewer packets are signed on all 84 rows: reviewer 1 Jino Gu,
reviewer 2 Jinghao Guo, each row carrying the independence attestation.
Reviewer 1: 52 needs_source, 32 approve_with_correction.
Reviewer 2: 45 approve, 32 approve_with_correction, 7 needs_source.

validate_human_reviews.py returns status=failed on two counts, both
recorded rather than worked around: review_date is written 2026/9/20
where ISO YYYY-MM-DD is required (168 rows), and all 84 items still
await adjudication. Signed rows were not edited. Gate 1 stays
in_progress; no item bank or response layer was rebuilt.

Adds ERR-20260920-043 and the reviewer summaries, including reviewer
2's findings on NSDUH source attestation, the GSHS French
mis-citation, census vs cluster design treatment, and the YRBS
1995/1997 Louisiana population caveat.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016yuKikAXbByLvsUnh55SRd
This commit is contained in:
Jinotech
2026-09-20 08:04:23 +00:00
co-authored by Claude Opus 5
parent 85f5dab88c
commit 413b679904
6 changed files with 340 additions and 168 deletions
Download Patch File Download Diff File Expand all files Collapse all files
research/stage1/Human Reviewer 2 findings.txt
+15
View File
@@ -0,0 +1,15 @@
The four NSDUH attempt rows can't be signed as they stand. Blank source_file, remote SAMHSA URLs, and the project's own audit admits those endpoints returned 403 and that page-level verification "is not claimed" — so nobody read them. But Dataset/NSDUH/2021NSDUHMRBWebSpecs101422.pdf is sitting right there with YSUI03 [IF CURNTAGE = 12 – 17] During the past 12 months, did you try to kill yourself?, which closes wording, population and skip logic immediately. What doesn't close: YUSUICTRY appears in no local dictionary or schema, yet the response dictionary asserts nine codes for it including 85/94/97/98/99 — unattested, and they look copied from YUSUITHK.
NSDUH ideation 2022–2024 has a codebook change-note ("Note: Beginning in 2022, questions YSUI01… were moved…") stored inside question_text. Never read to a respondent, and it feeds text-similarity linking. 2021 has the mirror-image defect — it drops a sentence the instrument actually contained.
The NSDUH language gap is already solved in your own data. LANGVER (1=English, 2=Spanish) is present for all four years, and the spec opens by asking the respondent which language they want. "Not established" understates what you have.
Six GSHS rows cite the French questionnaire but store English text (Morocco and Wallis and Futuna both ship English and French). Item numbers match in both, so it's a mis-citation, not wrong wording — but the cited file doesn't contain the stored text.
Jamaica and Wallis and Futuna were censuses, not cluster samples, yet survey_design.csv gives them identical weight/stratum/PSU Taylor treatment and marks it source_verified_fields.
Solomon Islands has no instrument anywhere under Dataset/ — searching "solomon" and "SLB" returns nothing. Its wording is inherited from the core GSHS form, not read from a Solomon Islands one.
YRBS 1995 and 1997 excluded a majority of Louisiana per their user guides; the population string claims plain national coverage.
The YRBS block otherwise held up well — all 51 wordings matched verbatim, and the variable numbering that looks off at a glance (2023: printed item 31, variable Q27) is correct in all seventeen years, checked against each guide rather than assumed. GSHS codes verified twice: Ghana's codebook reproduces cell-for-cell against GSHS.csv, and every component's row count matches its fact sheet.