On estimating cosmology-dependent covariance matrices

TL;DR

This paper introduces a statistical emulator framework to estimate the cosmology-dependent covariance matrix of matter tracer two-point functions using a limited number of simulations, aiding future cosmological survey analyses.

Contribution

It presents a novel method to predict cosmology-dependent covariance matrices with fewer simulations than traditional approaches.

Findings

The emulator accurately predicts covariance matrices across cosmologies.

The approach reduces the number of required simulations for covariance estimation.

It provides a foundation for planning efficient simulation campaigns.

Abstract

We describe a statistical model to estimate the covariance matrix of matter tracer two-point correlation functions with cosmological simulations. Assuming a fixed number of cosmological simulation runs, we describe how to build a `statistical emulator' of the two-point function covariance over a specified range of input cosmological parameters. Because the simulation runs with different cosmological models help to constrain the form of the covariance, we predict that the cosmology-dependent covariance may be estimated with a comparable number of simulations as would be needed to estimate the covariance for fixed cosmology. Our framework is a necessary first step in planning a simulations campaign for analyzing the next generation of cosmological surveys.