Reducing the impact of covariance uncertainties on statistical inferences of cosmological parameters

DANTE J. PAZ; ARIEL SANCHEZ

Garching

Workshop; Theoretical and Observational Progress on Large-scale Structure of the Universe; 2015

ESO, MPA, y Excellence Cluster Universe, TU Munich.

Today measurements of cosmological parameters on high precision cosmology are restricted at some extend by errors on the covariance matrix estimation. This is in particular the case of clustering measurements involving two-point correlation functions. For instance, in case of Baryon Acoustic Oscillation measurements (BAO), noise in the estimated covariance of the measurements can account for a 20% of the total error budget even when a prohibitive amount of mocks catalogues have been used on covariance estimation. In this talk we present a novel and general method for likelihood estimation based on a sort of weighing of the covariance matrix. Our approach exploits the Covariance Tapering Method (CTM) developed in the recent years by Kaufman el al. 2008. Our implementation reduces the effect of noisy large scale covariances, which are expected to have less statistical significance, whereas allow us precise measurements of likelihood confidence levels. Similar re! sults to those obtained by the standard technique are obtained using at least ten times less mock catalogues. When using the same amount of mock samples, the error in the precision matrix is reduced around 40% in comparison with the methods currently in use, whereas the confidence intervals approach even closely to the expected values. At the same time no bias has been observed on the target parameters. Finally we present some tests in the context of Gaussian random variables with realistic covariance and centred on a realistic model of the two-point BAO correlation function.Comments: