**Appendix B. Weights for Linking Participants in an RMO Key Comparison**

The contributions from linking participants in an RMO comparison are weighted to take account of correlations. With two linking participants, we follow ([6], Appendix B) and evaluate the required weights as a function of the uncertainties of

$$M\_{\mathbb{Z}} = \langle \overline{Y\_{\mathbb{Z}}} - \overline{Y\_{\mathbb{P}}} \rangle\_{A\_{\mathbb{Z}}} - D\_{\mathbb{Z},\prime} \tag{A3}$$

$$M\_{\rm P} = -D\_{\rm P} \tag{A4}$$

where P is the pilot (E in the initial CIPM comparison) and Z (I in the initial comparison) is the other linking participant. Writing the standard uncertainties in these quantities as *u*(*m*Z) and *u*(*m*P), and their covariance as *u*(*m*Z, *m*P), the weights are ([6], Equation (76))

$$w\_i = \frac{\frac{1}{\
u^2(m\_i)} - \frac{\mu(m\_Z, m\_P)}{\ln^2(m\_Z)\ln^2(m\_P)}}{\frac{1}{\
u^2(m\_Z)} + \frac{1}{\ln^2(m\_P)} - 2\frac{\mu(m\_Z, m\_P)}{\ln^2(m\_Z)\ln^2(m\_P)}}\tag{A5}$$

for *i* = Z and P. In this work, *ν*<sup>Z</sup> = 0.931 and *ν*<sup>P</sup> = 0.069 were obtained.

#### **References**

