Effect of data noise on the fit at stochastic variation of the true start vector. Five data points at seven concentrations were used to refit the CHA₁₆ model 1,000 times. Representative examples of successful fits are shown. Each parameter was stochastically varied 5-fold in both directions, resulting in a 25-fold total variation range. (A) Standard conditions with a noise level of 0.25. (B) Noise level 2.0. (C) No noise. (D) Incidence of negative parameters for the fits in A–C.