A critical reanalysis of supernova type Ia data

Abstract

Cosmological parameter fitting remains crucial, especially with the abundance of available data. While many parameters have been tightly constrained, discrepancies — most notably the Hubble tension — persist between measurements obtained from different observational datasets. In this paper, we re-examine the Pantheon supernova dataset to explore deviations in the distribution of distance modulus residuals from the Gaussian distribution, which is typically the underlying assumption. We do this analysis for the concordant cosmological constant model and for a variety of dynamical dark energy models. It has been shown earlier that the residuals in this dataset are better fit to a logistic distribution. We compare the residual distributions assuming both Gaussian and Logistic likelihoods on the complete dataset, as well as various subsets of the data. The results, validated through various statistical tests, demonstrate that the Logistic likelihood provides a better fit for the full dataset and lower redshift bins, while higher redshift bins fit Gaussian and Logistic likelihoods similarly. Furthermore, the findings indicate a preference for a cosmological constant model. However analysing individual surveys within the Pantheon dataset reveals inconsistencies among subsets. The level of agreement between surveys varies depending upon the underlying likelihood function.