W3 R95 and R95-A — original/independent exact observed-law ledger¶
For a plain-language account of R95 missing gaze and selective audit return, start with the R95 conditional identification article or the assumption-dependent missing-gaze bounds comparison. This ledger documents independent algebra and replay evidence; neither a returned-record audit nor this software identifies unreturned outcomes without external assumptions.
Scope: exact rational binary originally missing outcome identification, conditional on hypothetically externally known monotone direction/ratio of actual audit completion. There are no Monte Carlo replicates, participant observations, empirical validation, confidence intervals, or calibrated W3 simultaneous bands.
Original frozen exact design¶
| Evidence component | Immutable record |
|---|---|
| Original draft scientific source PR | #462 |
| Original head SHA | 1223015d70cb74c31e8f8cdeb0d088ccf0855b25 |
| Original scientific workflow | #39872191344, 4/4 PASS |
| Original immutable evidence | Artifact #12466408311, w3-r95-immutable-original-54-exact-directions |
| Original archived members | r95_exact_source.json, SHA256SUMS |
| Frozen design denominator | 3 invited nonreturn fractions × 3 known invitation rates × 6 directional/equality/ratio regimes = 54 deterministic exact rational cases |
| Public implementation/API status | Research-only, draft/unmerged, no changes |
Each original row has all four observed event probabilities, the exact sharp interval, its width, two attainable latent completion-law endpoints, an explicit external-assumption label, and no-release/no-coverage flags. The same observed law must hold for every hypothetical return direction.
Independent derivation and mandatory original archive replay¶
| Evidence component | Immutable record |
|---|---|
| Separate R95-A audit draft PR | #463, stacked on original #462 |
| Independent head SHA | edbe14c4a1003cc014325fd36124ada08b8c2184 |
| Independent workflow | #39872459692, 4/4 PASS |
| Independent evidence | Artifact #12467176301, w3-r95-A-independent-actual-original-54-case-replay |
| Mandatory condition | Download actual original workflow #39872191344 SHA256 archive; reject incomplete/tampered members; independently recalculate and match all 54 source records |
| Independent method | Allocate unknown unreturned \(Z=1\) mass \(t\), enforce response direction or positive ratio cap, verify both endpoint response witnesses and four-category observed event mass |
| Replay result | All 54/54 frozen source records reconstructed exactly from independent mathematical factorization |
| Claimed validation | Reproducibility of a narrow exact mathematical identification theorem, not real-world gaze sensitivity/coverage |
Analytical figure lineage¶
| Figure | Mathematical original | Displayed interpretation |
|---|---|---|
| R95 direction-conditioned sharp set | δ=1/100, \(q_0=q_1=(1-\delta)/2\), six frozen shape regimes | Sharp population mean endpoints (rounded), not confidence intervals |
| R95 width across return rates | δ=1/1000,1/100,1/20; widths \(\delta\), \(\delta/2\), \(\delta/[2(2+\delta)]\), \(0\) | Algebraic source widths; no confidence or Monte Carlo uncertainty |
Critical scientific restrictions¶
The latent response direction is not identified by the known randomized invitation probability or observed returned outcomes. A point-identified parameter under assumed \(c_1=c_0\) is not evidence that equality holds in nature. Equal observational likelihood across distinct latent laws is verified, so no method based solely on this observable event distribution can consistently distinguish those latent laws in general.
All original/audit research source remains draft/unmerged, the public package/API and stable 1.1.0 are untouched, and v1.2/GitHub/PyPI release is on hold. Scientific blocker #289 (simultaneous bands) and #296 (observation-dependent missing gaze) remain open. This article/figure package may merge documentation only, after 25/25 exact-head CI PASS and separate protected-main Pages docs-build/deployment PASS.