Same as Fig. 5 (HS6) but for a subset of the cosmological parameters (omitting WL alone) when transforming the contours back into the fundamental parameter |f_R0|. While in the logarithmic case the marginalised Fisher contours are simply Gaussians by definition, this is not the case for the contours on the linearly parameterised case |f_R0|. For the probes GC_sp (purple) and WL (blue) on their own which are less constraining, the errors on |f_R0| are very asymmetric with heavy tails on the right. However, the smaller the 1σ marginalised error on log₁₀|f_R0| is, the better the marginalised posterior on |f_R0| can be approximated by a Gaussian. This can be seen for the full combination of Euclid probes (yellow).