Distance Estimators' performance on Gaia DR2

Abstract

An improvement in the estimation of distance and distance modulus cannot be achieved by only an enhancement of the precision of the trigonometric parallax, but with the correct statistical treatment of the parallaxes to derive these parameters. We aim to provide a recommendation regarding the distance estimators to be used for Gaia DR2 and onwards, as well as to rise awareness about the practice of unquestioningly inverting the parallax. We test the performance of two Bayesian and a frequentist methods over a simulated sample of 107 Gaia DR2 stars, using a specifically developed Python software. We conclude that the use of the Bayesian method with the Exponentially Decreasing Space Density Prior improves the estimation of distance, since it has a good behavior for high relative error parallaxes, with a much smaller bias and dispersion than the rest of estimates.