Figure 6

Comparison of the robustness in parameter estimation. The right panels show the number of failed estimations (defined by > 5 fold difference with the true values) for S G and S I as functions of noise level. The dimensionless implementation is better at recovering the true values of the parameters. The right panels plot the ratio of the estimated to true parameter values for those after removing failed fittings. The data points at 0% noise was normalized to one.