Comparison of the impact of supernova type Ia calibrator samples (sets of hosts) on the Hubble constant inference H₀, each one relative to the baseline result, for the various light curve fitting codes. The triangle shows a case where for the bayesSN fitter only 52 out of the 55 hosts was calibrated due to several missing supernovae in the fit, and the cross one where only 32 out of 35 hosts were calibrated. (For details on the pie charts, see Fig. 13.) In this case we take 100% of the pies to be the case involving all supernovae data instead of the baseline.