Fitting Fragility Functions to Empirical Data