What would be the best way to measure and compare the variability of a variable displaying growth (or any other "main" variation)?
As a graphic example, let's suppose we have two subjects (blue and orange), performing a score Y which is improving across time. Although both are improving, their variability is different.
Would the residual mean error to the function best fitting for the central tendency be best? It should capture the variability, but it also depends on the goodness-of-fit of the function chosen to model the main tendency. I can image for more complicated functions as the linear model depicted here, this goodness-of-fit might also play a role and alter the residual mean error. Is there another way to capture this variability that is not affected by the goodness-of-fit?