Reproduction Kit: Posterior-Mean vs Best-Fit Diagnostics in w0CDM: Support Conditioning with Planck 2018 and DESI DR2 BAO

This dataset provides the reproduction kit accompanying the paper: “Posterior-Mean vs Best-Fit Diagnostics in w0CDM: Support Conditioning with Planck 2018 and DESI DR2 BAO.” Version 2 changes relative to v1: • Corrected PHANTOM best-fit minimizer result. The pre-committed curvature scan identified a lower-χ² region not reached by the original five-start minimization. Three scan-seeded rerun outputs are included, with two seeds converging to χ² = 1020.30 at closely matching parameter values. The original v1 PHANTOM minimum (χ² = 1021.20) is preserved as a historical artifact. • Added minimizer configuration files and starting-point generation scripts to document and audit the likelihood-dependent best-fit workflow for users with local Planck likelihood access. • Standardized conda environment naming across documentation and execution scripts. • Chain MCMC outputs are unchanged from v1. Contents of the archive: Chain-derived products (fully reproducible from this kit): • MCMC chains for three controlled w0CDM support configurations: CTRL2 (broad support), non-phantom (w0 > -1), phantom (w0 < -1) • Cobaya configuration files for the chain runs • Post-processing and figure-generation scripts • Posterior summaries, χ² block decompositions, normalized parameter shifts, and convergence diagnostics Likelihood-dependent archived products (regeneration requires local Planck likelihood access): • Best-fit minimizer outputs (original v1 result and v2 scan-seeded reruns) • Curvature-scan products along the (w0, H0) degeneracy direction • Minimize-mode configuration files and starting-point generation scripts Documentation included: • README.md with quickstart and full reproducibility instructions • REPRODUCIBILITY_MANIFEST.md with file-by-file provenance Reproduction pipeline: The default reproduce_all.sh script regenerates all chain-derived tables and figures from the archived chains using the included scripts and environment specification. Likelihood-dependent minimizer outputs are read from archived files. Regeneration of these products requires local installation of the Planck 2018 likelihood data and is documented in the README. The archive is intended to provide computational transparency and explicit provenance for all reported numerical values in the published analysis.