No evidence for missing covariance in the Pantheon+ SuperNova sample distance moduli

TL;DR

This study investigates whether unaccounted covariance could bias supernova-based cosmological measurements, finding no significant evidence for missing covariance and confirming the robustness of current supernova analyses.

Contribution

The paper introduces a Gaussian process-based method to test for missing covariance in supernova data, without assuming specific physical models, and applies it to Pantheon+ samples.

Findings

No statistically significant missing covariance detected.

Upper limit for covariance amplitude is 0.031 mag at 95% confidence.

The Hubble tension significance reduces from 5.3σ to 4.5σ.

Abstract

Inspired by the discussion in the community on possible hidden systematic errors in late universe cosmological probes and non-trivial physical models developed to reduce the Hubble tension, we investigate the Pantheon and Pantheon+ SNe samples for possible deviations from the original $\Lambda$CDM analysis. To simultaneously account for possible systematics or deviations from $\Lambda$CDM, we adopt Gaussian processes to model additional covariance while making no further assumptions on their origin. We explore both stationary and non-stationarity corrections to the covariance. While small changes in the inferred cosmological parameters $H_0$ and $\Omega_{m}$ can occur, we find no statistically significant evidence for missing covariance. We find an upper limit for the Gaussian processes amplitude $\sigma < 0.031$ mag with $95\%$ confidence, which corresponds to $20\%$ of the average statistical error in the Pantheon+ sample. The strongest effect we find on the inferred cosmological parameter posterior can reduce the statistical significance of the Hubble tension between Pantheon+ and Planck estimates from 5.3$\sigma$ to 4.5$\sigma$. Therefore, we conclude that the SN cosmological parameter inference is robust against the analysis modifications studied in this work.