Hi all,
I left this ROC feat until last.
Am confused why the total number of true toxic patients for the KPS score (sum of toxic patients correctly predicted by each KPS strata, i.e. 128+76+53 = 257) doesn't equal the total number of true toxic patients for the chemotherapy risk score (39+117+90 = 246). Shouldn't they be the same, i.e. a fixed proportion of n = 500 will have true chemotherapy toxicity regardless of whether KPS or chemo risk score predicition is applied?
Perhaps this is just due to my own cognitive failure and I have really misunderstood the table. But I can't see how these totals could be different.