Effects of initial values of coefficients on automated curve fitting. A 10% level of Gaussian noise was added to a bi-linear model velocity profile to create a "Monte-Carlo-dataset" (gray dots, n = 1301; coefficients are b1 = 0.1, b2 = 1.84, c1 = 0.3, c2 = 0.8, d1 = d2 = 50), and the step-stool equation was fitted using three different sets of coefficient values to initiate the fitting algorithm. The graphs of three of these sets are shown in (A); the blue and the green curves are deliberately inappropriate, whereas the red one represents a "smart guess". The result of automated curve fitting starting from the initial values depicted in (A) is shown in (B) in corresponding colours. In the cases of poor initial value choice, the fitting algorithm became stuck at unacceptable solutions; in contrast, the "smart guess" provided a satisfying result.