{
  "harness": "sj-harness 0.6.0",
  "job": "job:2bf7db1f536f1e95faa00c427af5b8f9",
  "kind": "domain_review",
  "bundle": "sha256:25c9902bacc8de4838daa29311af0df8c06aaf611ac1d5cc759c4312cfbfd660",
  "claims": [
    {
      "local_id": "C1",
      "claim_id": "claim:572af7b42c7e72b73c70aa811a48ffef7ce1bc1dd1dd277ace5ba95804a5f59d",
      "verdict": "sound",
      "significance": "minor",
      "reason": "Numbers match S.json and agree with Foster and Rahmstorf's reported over-99% changepoint confidence in 2013/2014; the port reproduces the published adjusted series. Minor note: FR26 Table 1 gives changepoint times by month while this test uses annual means."
    },
    {
      "local_id": "C2",
      "claim_id": "claim:dd25c33bfcf43a7834074478211b7dfc6e77f05c66d502f854034b959e5d30d2",
      "verdict": "sound",
      "significance": "moderate",
      "reason": "A proper exact Monte Carlo test of the stated null, re-estimating the whole adjustment in each record; size 5.9-7.0% and p 0.0005-0.0025 support robustness. It answers the unpropagated-adjustment critique (RealClimate 2026; Bennedsen and Vera-Valdes 2026) with a number; the fixed-exogenous-coefficient limitation is stated."
    },
    {
      "local_id": "C3",
      "claim_id": "claim:492cb86dec1284842dad5beebf27f42d4d5fcc6d82527a933cedc236b8032c72",
      "verdict": "major_issues",
      "significance": "minor",
      "reason": "Size and two-sided p reproduce from null_draws.csv, but FR26 does not say whether its 98% is one- or two-sided, or monthly or annual. One-sided full-pipeline p from the bundle's own draws are 0.0005-0.017 on annual means (all below 0.02) and above 0.02 only for HadCRU monthly (0.024), so the 'does not hold in all five' conclusion depends on an undiscussed sidedness choice. Over-rejection (7-9% at nominal 5%) stands; Glaser, Richardson and Landerer 2024 is relevant uncited prior work."
    },
    {
      "local_id": "C4",
      "claim_id": "claim:56fe710433c53da4abda0e8397247c1051602882ada2570a7028661d775deba2",
      "verdict": "minor_issues",
      "significance": "minor",
      "reason": "Numbers match S.json and it is properly framed as a sensitivity analysis. GPH with 25 frequencies gives d an SE of about 0.13, which is not reported, and the long-memory GMST literature (Mann 2011; Rypdal and Ostvand 2013; Vera-Valdes 2021) is not engaged."
    }
  ]
}