Modeling the probability distribution for cosmological analysis with photometrically classified samples

TL;DR

This paper introduces a simplified probabilistic model for incorporating photometrically classified supernovae into cosmological analyses, improving constraints on cosmological models compared to traditional methods.

Contribution

It proposes a new likelihood approach that models contamination as a redshift-dependent mean shift, outperforming the BEAMS framework in supernova cosmology.

Findings

The new model is strongly favored by Bayes factor across configurations.

It enhances the constraining power of photometric supernova data.

Comparison with existing classifiers shows consistent improvements.

Abstract

In this work we investigated methods for the accurate and efficient incorporation of photometrically classified supernovae into cosmological analyses, and to assess the impact of the additional uncertainty associated with this procedure on the ability of Type Ia supernovae (SNeIa) tests to place constraints on cosmological models. We proposed a simplified likelihood, in which the contamination is described as a redshift dependent change in the mean of the usually assumed Gaussian distribution, and we tested this hypothesis against the usual two-component approach, based on the BEAMS framework. Using the latest version of the DES supernova sample, dubbed DES-Dovekie, we compared the results when using type probabilities from different classifiers, such as SNIRF and SCONE, and applying different cuts on these probabilities. We show that the new model is strongly favored by the Bayes factor, when compared with the current one, for all configurations, allowing an improvement on the constraining power of photometric supernova data.