An ROC curve is plotted with (1-False Positive Rate) on the X-axis and the True Positive Rate on the Y-axis. However, the way in which each point of the curve is plotted is by first picking a cut-off probability value below which all samples are classified as negative and above which all are positive. And using this decision threshold, we get one point (TPRpoint 1, 1-FPRpoint 1). Doing this for multiple decision thresholds/cut-off probabilities, we can generate a 2-Dimensional ROC curve.
Does this mean that the true independent value is the cut-off probability and not the FPR as looking at the X-axis would suggest?
