Pedrotti/de Chambrier sampling + missingness — authoritative result

Status

This document records the first scientific execution of the frozen Pedrotti/de Chambrier sampling-and-missingness protocol. The outcome is authoritative for this case study. It was executed only after the protocol, endpoint-blind public source freeze, independent source-lock binding, scientific implementation, and archive-before-reveal workflow had been fixed and certified.

No scientific choice in this record was selected or changed after observing the outcome.

Immutable provenance

The exact critical execution environment was Python 3.12.14, NumPy 2.5.3, and pandas 2.3.3.

Scientific archive

The workflow verified the scientific payload, uploaded the artifact, and only then executed the reveal step.

The archive SHA256SUMS was independently verified and every declared file matched. The GitHub-provided artifact digest and independently recomputed ZIP digest are identical. All 165 individual recovery diagnostics were recomputed from the archived estimates with zero mismatches, and all nine family-recovery fractions were reproduced.

Frozen estimand

The endpoint is the participant-balanced long-minus-short numeric gaze-path-rate difference, expressed in pixels per second.

For each participant, trial-level path displacement is divided by the original trial duration. Short and long numeric trial means are formed within participant, their difference is computed, and participant contrasts are averaged without weighting at the study level.

Native missingness is preserved. Path displacement is accumulated only across consecutive scheduled rows whose two gaze endpoints are finite; gaps are not bridged.

Reference result

Reference estimate: -75.37814047218419 pixels/second.

The negative direction means that, under this frozen endpoint definition, the long numeric condition had a lower participant-balanced gaze-path rate than the short numeric condition.

Sampling-rate sensitivity

Sampling-family recovery is therefore 0.4.

Downsampling preserved the negative sign at all five target rates, but 125, 100, and 50 Hz exceeded the predeclared 20% relative-deviation tolerance.

Added-missingness sensitivity

Family recovery fractions were:

Each missingness family contains the 20 prespecified deterministic replicates generated from root seed 20260913.

Authoritative classification

Classification: materially_fragile.

The frozen rule requires strict sign agreement plus no more than 20% relative deviation for an individual estimate. A family is robust only when its recovery fraction reaches the prespecified minimum; the full execution is classified materially_fragile when any family reaches the material-fragility ceiling.

The independently recomputed family table contains several recovery fractions at 0.4 or 0.0, so the predeclared rule necessarily returns materially_fragile.

This classification is about recovery of the magnitude-and-direction criterion. It should not be paraphrased as sign reversal: every sampling-rate estimate remained negative.

Inferential boundary

This case study is a robustness/sensitivity audit. It does not provide a population confidence interval, posterior interval, p-value, or causal effect claim. Its purpose is to show how strongly the frozen descriptive endpoint depends on sampling resolution and controlled missingness perturbations.

Post-outcome scientific freeze

The following may not be changed in response to this result: source cohort; source-file identity; endpoint or contrast direction; dominant-eye/native-timeline policy; sampling grid; target rates; missingness mechanisms; missingness fractions; replicate count; seed derivation; original-duration denominator; 20% relative tolerance; strict-sign rule; family recovery threshold; material-fragility ceiling; or final classification logic.

Permitted follow-up is limited to independent artifact verification, deterministic reproducibility checking without selecting among outcomes, archival/provenance work, publication reporting, and non-scientific infrastructure maintenance.