We present a comparative analysis of estimators and Bayesian methods for determining the number count dipole from cosmological surveys. The increase in discordance between the number count dipole and the CMB’s kinematic dipole has presented a challenge to the assumption of an isotropic and homogeneous universe. The level of discordance has depended on the choice of method to determine the dipole; hence, there is a need to compare them to determine the better approach. We select the healpy fitdipole algorithm as our estimator and show that it gives unbiased results regardless of the noise levels and sky coverage in the data. However, for low sky coverage, the estimator’s results have a large variance, which indicates that the estimator is not reliable in that regime. We then compare the estimator’s outputs with the Bayesian results. If the sky coverage is sufficient, the Bayesian posterior probabilities agree with both the estimator’s outcomes and the true parameter values. But if the sky coverage is low, Bayesian analysis is often inconclusive, which safeguards against incorrect conclusions. Both methods provide the ability to analyse multiple samples: on one hand, we need to select a different estimator, while in Bayesian inference, we need to extend our likelihood function to incorporate additional parameters. Our study emphasises the need to perform a principled statistical analysis of sky surveys for dipole determination.
Mittal et al. (Mon,) studied this question.