Answers are grounded in this paper and cite their sources.
Pre-analysis question register (v1.0) for a test of the Helm-failure hypothesis in the Robodebt case, using existing CAMS records. It is a question register, not a completed statistical analysis plan. Deposited before any pre-specified analysis; it is not a claim that the investigator had no exposure to the data. The accompanying experimental record (experimental_record_20261113.txt) documents the same-day scoring sessions, including that CHRPLOT trajectory plots of CES/Centrelink (6-month) and Australia (quarterly) series were generated and delivered, that scoring used event overrides (including a Robodebt event) and deterministic jitter, and that blinded sets were created afterwards. These facts bear on blinding, leakage and provenance (register items Q31-Q33) and limit claims to retrospective discrimination. The SHA-256 digests of the register, the Australia scoring files (codebook v1.4, panel-build script) and the experimental record are in the three manifests; each manifest carries an OpenTimestamps proof (.ots), pending Bitcoin confirmation until upgraded. The register's access statement is the investigator's report and has not been independently certified. Scoring data files are not deposited here.
No takes yet. Share an insight, caveat, or question.
Kari McKern (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: