S5 study packet - Honest Government Research

Everything behind the locked S5 result, as one coherent set: the pre-registration, the S5 methodology brief, the frozen data, the fielded instrument, and the analysis code that turns the data into the published numbers.

Start here

The pre-registration and the methodology rationale for the locked wave.

S5 pre-registration

The plain-English pre-registration summary: design, exclusion rules, and what each test means (filed on OSF before analysis).

Open

S5 methodology rationale

Why the S5 study is built the way it is, and how it avoids leading respondents. True for S5; does not borrow S6 design claims.

Open

Data

The frozen responses and the published numbers. These reproduce every figure on the site.

data/rthg-s5-dataset.jsonl

The frozen, de-identified responses, one per line. No Prolific IDs (salted pseudonyms only); raw written answers removed.

918.7 KBDownload

data/rthg-s5.json

The published numbers, the dataset fingerprint (SHA-256), and the freeze date.

100.2 KBDownload

data/rthg-s5-figures.json

The exact figures rendered on the public site.

1.5 KBDownload

Survey instrument

The authoritative wording, item structure, and the S5 exclusion rules as fielded.

lib/rthg-study-definition.ts

Every page, question, scale, and principle id, exactly as fielded, plus the analysis-plan constants.

115.4 KBDownload

lib/rthg-study-flags-s5.ts

The pre-registered S5 exclusion and data-quality flag rules (B-rev excluded at 3 or higher; vote-2024 citizenship proxy).

9.0 KBDownload

Analysis code

The script and engine that turn the raw data into the published numbers. The same code the /verify page runs in your browser.

scripts/rthg-reproduce.ts

Recomputes every published figure from the raw data and prints MATCH or MISMATCH against the published values.

5.6 KBDownload

lib/study-analysis/index.ts

The analysis engine entry point: ties parsing, filtering, and statistics together.

1.7 KBDownload

lib/study-analysis/parse.ts

Normalizes a raw response row into the engine’s internal shape.

3.7 KBDownload

lib/study-analysis/filters.ts

Applies the pre-registered exclusion rules and builds the sample funnel.

2.9 KBDownload

lib/study-analysis/stats.ts

The pure statistics (support shares, Wilson intervals, principle agreement, party breakdown).

9.6 KBDownload

lib/study-analysis/waves.ts

Per-wave field keys and page ids (taken verbatim from the survey instrument).

7.2 KBDownload

lib/study-analysis/types.ts

The data structures used across the analysis engine.

5.2 KBDownload

Unit tests

Every rule in the analysis engine has a unit test. The same tests run inside the /verify page (in your browser) and on the command line. A green suite means the deployed code matches the documented behavior.

lib/study-tests/framework.ts

Tiny framework-agnostic test runner used by both vitest and the in-browser runner.

5.8 KBDownload

lib/study-tests/fixtures.ts

Synthetic clean-baseline row builders for S5, S6, and s6-short. Tests perturb one field at a time to exercise a single rule.

6.0 KBDownload

lib/study-tests/parse.tests.ts

Reads vote, mechanism, panel, and demographics from the right keys per wave; B-rev reverse-scoring; no-opinion handling.

6.1 KBDownload

lib/study-tests/stats.tests.ts

Wilson 95% CI, mean, median, vote stats, principle stats, mechanism, shift, post-stratification.

12.1 KBDownload

lib/study-tests/filters-s5.tests.ts

Every S5 frozen exclusion rule (attention, reverse-coded, comprehension, method, speed, final check, no-opinion).

6.5 KBDownload

lib/study-tests/cell-assignment.tests.ts

Seed determinism, cell assignment, Section B fixed positions, and a 1,000-seed property test of the no-three-in-a-row cluster rule.

7.4 KBDownload

lib/study-tests/waves.tests.ts

Each wave config has the right arm prefix, mechanism rating prefix, panel page reuse, and post-strat benchmark.

3.9 KBDownload

lib/study-tests/cohorts.tests.ts

Cohort funnel narrows monotonically; vote-2024 proxy applies on S5 only; non-citizen flag drops on S6+.

5.3 KBDownload

lib/study-tests/run.ts

Command-line entry that runs every test outside vitest and exits non-zero on failure.

800 BDownload

lib/study-tests/__vitest__/all.test.ts

Vitest adapter so the same tests report in a standard test runner.

570 BDownload

AI audit prompt

The copy-paste prompt for the no-code AI verification path.

AI_AUDIT_PROMPT.txt

The copy-paste prompt for the no-code AI verification path (S5).

2.4 KBDownload

Download the S5 packet as ZIP Verify the result yourself →Run your own replication