AOMIC ID1000 replication — completed 27 October 2026

The original internal plan, receipt and source manifest are unchanged.
The source dataset is public CC0: OpenNeuro ds003097 v1.2.1,
DOI 10.18112/openneuro.ds003097.v1.2.1.
Cite Snoek et al. (2021), Scientific Data 8,85,
https://doi.org/10.1038/s41597-021-00870-6.

Reproduce in a copy of this folder to preserve the supplied run timestamps:
  python -m pip install -r requirements_AOMIC.txt
  python fetch_sources.py
  python run_locked_replication.py
  python independent_verify.py

The fetch script downloads four public source files at a fixed commit and
checks SHA256 values. The full public participant table includes irrelevant
variables and is downloaded on demand rather than duplicated here.
The minimal analysis data contain only study variables and derived scores.
Preserve the original analysis lock. Do not create a new lock retrospectively.
Future extensions must be separately dated and distinguished from this plan.

Primary N=926; global F(4,918)=3.1444774132, p=.0139590003;
partial R²=.0135162349. Focal standardized beta=.0968349999,
conventional 95% CI [.0241733152,.1694966845], p=.0090569746.
HC3 inference and 2,000 bootstrap resamples support the primary result.
All nine frozen sensitivities are retained. The eTIV-only sensitivity is null;
the result is specifically conditional on total analyzed tissue volume.

Independent verification means a separate computational implementation, not
independent human review. AOMIC does not correct HCP family dependence.
