I have 10 different curves, and the curves are on different ranges. For example, curve 1 has range [-1,1], curve 2 has range [-10,10], etc.
Now based on my model, I get one estimated curve for each of the 10 curves, so that I can calculate the MSE for each of the 10 curves.
Now the problem comes. Since the curves are in different ranges, it is not fair to compare the MSE directly, right? So I wonder anyone has any idea on developing a new metric that can take into account the different ranges?