Figure 1
Non-weighted nonlinear least squares Hill model fits to 5 datasets. Residuals of Munch-Petersen et al. 1993 and Berenstein et al. 2000 show a trend from positive to negative values across lower fitted values, i.e. poor fits. The dataset of Birringer et al. 2006 is an outlier in that its kmax is 15-30-fold smaller than those of the other datasets. The data of Li et al. 2004 is different in that its Hill coefficient is ~1 rather than 1.24-1.30. The 2004 data show that the variance increases with increases in fitted values, as it should since activities cannot be negative and thus the activity variances must decrease with decreasing expected values. The Hill coefficients presented here approximately equal those of the original publications.