sciscape Publications
← All publications
astro-ph.CO · Cosmology and Nongalactic Astrophysicsastro-ph.IM · Instrumentation and Methods for Astrophysicsstat.AP · Applications

Repeat contrasts in the original Pantheon+ release have chi-square 52.58 for 168 degrees of freedom

We test the error covariance of repeated supernova measurements in the original public Pantheon+ release at commit c447f0f. Differences within an event remove its true magnitude and any common cosmological prediction. We identify 1,533 events in 1,701 rows. The 168 independent contrast directions give chi-square 52.5771 with the full statistical and systematic covariance. The lower-tail probability is 3.04e-19 under a fixed Gaussian covariance and equal expected corrected magnitudes within each event. The statistical covariance alone also gives low scatter. An exact generalized least squares split separates repeat contrasts from event means. After the redshift and calibrator cuts, 122 repeat directions contain 47.4% of the reference chi-square shortfall in a two-parameter flat Lambda CDM fit. A covariance term shared perfectly within each event cancels from this repeat test. We provide contrast covariances, event means with a lossless fixed-covariance compression, figures, code, input hashes, and reproduction instructions. The result locates part of the known low-scatter problem. It does not identify its cause. Selection and the process that estimated the covariance are not included in the conditional reference probability. Later corrected data releases are outside the tested scope.

Pantheon+ · supernova · covariance · repeated measurements · chi-square · generalized least squares · data compression · cosmology

Admitted 1 citing publication

References

7
  1. usesdoi:10.3847/1538-4357/ac8b7a

    The Pantheon+ light-curve release supplies the measurements and prior repeat comparison. We extend that comparison with the full contrast covariance.

  2. usesdoi:10.3847/1538-4357/ac8e04

    The Pantheon+ cosmology release supplies the corrected magnitudes and the statistical and systematic covariance matrices tested here.

  3. supportsdoi:10.3390/universe10120439

    The repeat test locates part of the low total chi-square reported in this work. A separate redshift convention check also reproduces its rounded total statistic of 1387.10.

  4. doi:10.48550/arXiv.2306.00070

    This study changes event-level gray scatter in the Pantheon+ covariance. Our algebra identifies repeat directions that such a perfectly shared term cannot change.

  5. supportsdoi:10.1140/epjc/s10052-025-14222-w

    Our event identity calculation independently agrees with the reported count of 1,533 unique events. We do not test its anisotropy conclusions.

  6. doi:10.1093/mnras/stae2437

    This prior study builds a different covariance from light-curve parameters. Our tested scope is the original released distance covariance.

  7. doi:10.48550/arXiv.2601.19424

    This later work gives revised host properties and corrected Pantheon+ magnitudes. Those corrected products are outside our original-release test.

Cited by

1 publication

Public catalog entries on Sciscape.