Bayesian angular power spectrum analysis of interferometric data

TL;DR

This paper introduces a Bayesian inference method using Gibbs sampling for analyzing interferometric data to estimate the angular power spectrum and signal maps, applicable to cosmic microwave background and 21 cm observations, demonstrating computational efficiency and adaptability.

Contribution

The paper develops a Gibbs sampling-based Bayesian framework for angular power spectrum inference from interferometric data, accommodating arbitrary uv-plane coverage and heteroscedastic errors, suitable for future cosmology missions.

Findings

Validated with mock CMB data in the flat-sky approximation

Computational complexity scales as O(n_p log n_p)

Demonstrates adaptability to various interferometric observations

Abstract

We present a Bayesian angular power spectrum and signal map inference engine which can be adapted to interferometric observations of anisotropies inthe cosmic microwave background, 21 cm emission line mapping of galactic brightness fluctuations, or 21 cm absorption line mapping of neutral hydrogen in the dark ages. The method uses Gibbs sampling to generate a sampled representation of the angular power spectrum posterior and the posterior of signal maps given a set of measured visibilities in the uv-plane. We use a mock interferometric CMB observation to demonstrate the validity of this method in the flat-sky approximation when adapted to take into account arbitrary coverage of the uv-plane, mode-mode correlations due to observations on a finite patch, and heteroschedastic visibility errors. The computational requirements scale as O(n_p log n_p) where n_p measures the ratio of the size of the detector array to the inter-detector spacing, meaning that Gibbs sampling is a promising technique for meeting the data analysis requirements of future cosmology missions.