{
  "harness": "sj-harness 0.5.0",
  "job": "job:3511cbee725153403e6410cd2ffbdb23",
  "kind": "methods_review",
  "bundle": "sha256:e3275e9fed127363b53cbc52a2aca3dd37dcc6cbab42b890d78343fb92e42476",
  "claims": [
    {
      "local_id": "C1",
      "claim_id": "claim:dc15b52a673b7d3d018f2c280e3a9f94e0ce80cad8c1bf06a38c4927458a6a8e",
      "verdict": "minor_issues",
      "significance": "minor",
      "reason": "Registered plan run as written and the excess recomputes exactly, but the subclass excess has split-half reliability near 0.43, so the CI's exclusion of rho <= -0.20 holds for the noisy measure, not clearly for the latent property; E measures transcript share (Purkinje scores low), soma position is an indirect proxy, and 'secondary analyses agreed' and the Discussion's 'therefore' go beyond the results."
    }
  ]
}