A pre-registered identifiability test for a common latent state in an optical clock frequency-ratio network

This record contains the manuscript and the complete reproduction material for the study “A pre-registered identifiability test for a common latent state in an optical clock frequency-ratio network”. The study tests whether an additional common latent state is identifiable in a heterogeneous optical-clock frequency-ratio network after accounting for the admitted node and architecture structure. The analysis was conducted under a pre-registration that fixed the hypothesis space, comparison metric, admissibility conditions, diagnostics, decision margins, and admissible concluding statements before any likelihood was evaluated. The composed data cloud contains 132 measurement edges over 10 network nodes, representing 23 physical comparison identities, 109 states, and 18 architecture groups. The predictive comparison uses 13 admissible masked-edge cross-validation folds. Of 69 registered model members, 54 were scored on every fold. The 15 members of the linear-loading class were not scored because their fitted loading structure did not satisfy the pre-registered certification and diagnostic requirements. No scored latent-state member satisfies the pre-registered superiority rule. The largest improvement over the null model is +0.390, below the pre-declared margin of 2, whereas the architecture-flexible benchmark criterion is satisfied. Under the pre-registered decision language, the applicable conclusion is: H₀ is sufficient; no additional common latent state is identifiable. This conclusion is restricted to the composed data cloud, admissible folds, hypothesis space, covariance assumptions, and decision rules reported in the manuscript. Reproduction archive The accompanying archive PhaseA_Reproduction_Repo_v0p2_ZENODO.zip contains the frozen inputs, pre-registration and protocol records, authoritative analysis engine, run outputs and diagnostics, historical comparison material, provenance and governance records, and reproduction tools used for the reported analysis. The repository is version 0.2. Every repository data member carries its SHA-256 digest in the deposited inventory. Internal checksums provide an additional integrity check for the archived repository. Container integrity SHA-256 of PhaseA_Reproduction_Repo_v0p2_ZENODO.zip:f37a5da38c9c9a1efa58ad5f1d87f8dfaedb08309c7fae8bf5d5e575bd8b3ca1 Source data and licensing The archive contains 02_inputs/S2_fresh.zip, redistributed unchanged and byte-identically from the PTB dataset published at DOI 10.7795/720.20210126. The licensing terms of that source dataset, including its CC BY-ND 4.0 licence, continue to apply to that redistributed material. The CC BY 4.0 licence assigned to this Zenodo record does not supersede or replace the licence applicable to that third-party material. The frequency-ratio campaign data used in the analysis are those published at DOI 10.1364/OPTICA.561754. Contents of this record The record contains:– the manuscript;– the manuscript source;– the Phase-A reproduction repository, including the scientific inputs, analysis implementation, frozen outputs, diagnostics, provenance records, and reproduction tools. The manuscript and reproduction material are deposited together in this record. The Zenodo DOI of this record therefore identifies the archived record as a whole; it is not used as a self-reference in the manuscript’s Data Availability statement. Versioning The manuscript deposited in this record is manuscript version 0.25. The reproduction repository is version 0.2. The accompanying supplement is version 0.1 where included in the reproduction material. These version designations refer to their respective artefacts and should not be conflated.