Output distributions of the naïve (black), MAS (orange), and the MB (pink) estimators applied to 100 000 Monte-Carlo simulations using a covariance matrix in the canonical case (ε_eff = 1), for three levels of the S/N p₀/σ_p = 1,2, and 5 (from top to bottom). In the case of the MB estimator, we show two setups of p₀ = 1% and 50% to illustrate the dependence of the output distribution on the p₀ value, due to the prior used in the Bayesian approach ( so that ). The other estimators are not sensitive to the true value p₀. The MB and MAS curves overlap in the bottom panel.