Weighted nonlinear least squares Hill model fits. Reciprocal data squared weights of 1/k2 were used to minimize the sum of squared relative residuals. Compared to Fig. 1 the Hill coefficients here strike a better balance between low and high substrate concentrations. The Hill coefficient of Li et al. 2004 is now similar to those of the 1993 and 2000 datasets. The Hill coefficient of Frederiksen et al. 2004 is now an outlier at h = 1.6.