Model accuracy
•
The figure shows how the performance of an NTCP model deteriorates as
confounding factors are increased in the patient population to which the
model is fitted
•
The area under the curve (AUC) is a measure of accuracy, giving the
probability that the model will correctly rank sampled ‘complicator’ & ‘non-
complicator’ pairs from the data set
– 0.5-0.7: poor
– 0.7-0.9: moderate
– 0.9-1: excellent
05/04/2016