PDF of residuals for predictor P_i = Emu_i and the residual based on predictions from the average values (P_i = ⟨Y⟩), with respect to the missing data point y_i. Data are computed on overdensity Δ_200c. The upper row shows the PDF for y_i = A_i (normalisation), and the lower row shows the PDF for y_i = β_i (log-slope). The scatter obtained with the emulator is significantly smaller than the residuals on average. Grey lines correspond to the residuals when the emulator is trained without either σ₈ (dotted line), h₀ (dashed line), or when using Ω_b/Ω_m instead of training it with Ω_b and Ω_m separately (dash-dotted line).