Figure 2

Logistic regression model output for each tumor grade. The plots on the left show the logistic regression model probabilities (P) using the age and grade variables as the model inputs for each tumor grade. The plots on the right show the respective hybrid logistic regression model probabilities (P) using the variable z (i.e. age and grade combined with the probabilistic neural network) as the model input. Because there are overlapping points (patients with the same grade and age), some points are not distinguishable. The censored group (black) is compared with the incident group (red). The curves were fitted with a cubic spline.