Bias in the correlation functions in the three-sample case of Section 5.2, with 1σ uncertainties in each estimator indicated with the shaded regions. We see that as in the toy example in Section 5.1, just as ﹩{\mathtt{Decontamination}}﹩ (red) reduces the bias using the correlations on the contaminated subsamples (blue), the Decontaminated Weighted estimator (black) reduces the bias from the Weighted correlations on the entire observed sample (magenta), without needing to divide the observed sample into subsamples; the green line indicates zero bias.