Fig. 2
From: Emulating complex simulations by machine learning methods
Dynamics of the true output variable versus the fitted one together with the \(95\%\) bootstrap confidence interval. On the x-axis there is the time of the trajectories in weeks while on the y-axis the output variable normalised to 0–1. The first panel from the top that refers to \(\beta (t)\) shows a good fitting precision as the true trajectory always falls inside the confidence band. The second panel refers to \(\gamma (t)\) and it shows that again the emulator shows good fitting performance The third panel refers to \(\tau (t)\) and shows that while the emulator is able to catch the overall trajectory, due to the high variability of the true trajectory, the smoother confidence band does not embed it completely