Correlating uncertainties in global analyses within standard model EFT matters

We investigate the impact of correlations between (theoretical and experimental) uncertainties on multi-experiment, multi-observable analyses within the standard model effective field theory (SMEFT). To do so, we perform a model-independent analysis of $t$-channel single top-quark production and top-quark decay data from ATLAS, CMS, CDF, and D0. We show quantitatively how the fit changes when different experimental or theoretical correlations are assumed. Scaling down statistical uncertainties according to the luminosities of future colliders with $300{\mathrm{fb}}^{-1}$ and higher, we find that this effect becomes a matter of life and death: assuming no correlations returns a fit in agreement with the standard model while a “best guess”-ansatz taking into account correlations would observe new physics. At the same time, modeling the impact of higher order SMEFT-corrections the latter turn out to be a subleading source of uncertainty only.

Figure 1

Marginal constraints on the coefficients ${\tilde{C}}_i$ from fits with the no correlation scenario, Eq. (7), and the best guess scenario, Eq. (8). Dots and lines denote the central value and the smallest 95% interval, respectively, in the 1D projection. The SM is indicated by the vertical dashed line.

Figure 2

Central values of ${\tilde{C}}_{\varphi q}^{(3)}$ in the different EFT-implementations for correlation parameters ${\rho }_{\mathrm{sys}},{\rho }_{\mathrm{th}}=0.0$, 0.3, 0.6, 0.9 from a marginalized fit to the data given in Table 1. Both correlation parameters are varied independently from each other. The upper-left and lower-right corner correspond to the no correlation scenario in Eq. (7) and the best guess scenario in Eq. (8), respectively. Central values below and to the right of the grey line are in conflict with the SM at more than $2\sigma$.

Figure 3

Central values (red line) and 95% intervals (red band) of ${\tilde{C}}_{\varphi q}^{(3)}$ in the different EFT-implementations for correlation parameters ${\rho }_{\mathrm{sys}}={\rho }_{\mathrm{th}}\in [0,0.9]$ from a marginalized fit to the data given in Table 1.

Figure 4

Histograms of the central value and the size of the 95% interval of ${\tilde{C}}_{\varphi q}^{(3)}$ in the linear EFT-implementations for correlation parameters varied randomly around the best guess scenario from a marginalized fit to data given in Table 1. Black lines denote results from the best guess scenario in Eq. (8).

Figure 5

Same as Fig. 1, but with statistical uncertainties of the data in Table 1 scaled to $300{\mathrm{fb}}^{-1}$, assuming present central values, systematic uncertainties and theory uncertainties.

Figure 6

Same as Fig. 2, but with all statistical uncertainties of the data in Table 1 scaled to the expected integrated luminosity of $300{\mathrm{fb}}^{-1}$, assuming present central values, systematic uncertainties and theory uncertainties.