Fig. 3.
Model goodness of fit. (A) The figure shows the error in NRS prediction (y axis) as a function of the predicted NRS score (x axis) for the population fit, reflecting the effects of covariates known to the model. The red line is Friedman SuperSmoother. The shaded regions show the portions of the figures where prediction errors cannot occur because both measured and predicted NRS are bounded by 0 and 10.(B) The figure shows the error in NRS prediction (y axis) as a function of the predicted NRS score (x axes) for the individual fit, reflecting the effects of covariates known to the model as well as random intersubject differences estimated by NONMEM (NONlinear Mixed Effects Modeling program). The red line is Friedman SuperSmoother. The shaded regions show the portions of the figures where prediction errors cannot occur, because both measured and predicted NRS are bounded by 0 and 10. The fit shows how well the model could work, given the limitations of the measurements and the model structure, if all of the intersubject variability could be explained by patient or treatment characteristics. NRS = numerical rating score for pain.