Dear All,
One of the ways to find the best cut-off value on an ROC curve is to find a point on the curve with a certain slope, say 'a'. My question: How can we find such a point in practice while we do not have the curve---in fact, we have only few points (discrete data) obtained from our experiment. To many data set from practical data, say medical experiments, we can fit numerous curves (many of them with acceptable fitting) to the data set. Then, it is not easy to find the slope at a given point or the point where the slope is 'a'.
Thanks,
Farrokh
