ROC

 

The ROC tab contains the following elements:

•

ROC Plot

This overlay plot is the standard Receiver Operating Characteristic curve (ROC), plotting sensitivity versus 1 - specificity. The area under this curve is a measure of the sorting efficiency of the predictive model, with 1 representing perfect sorting and 0.5 random chance. Better curves come closer to the upper left corner of the plot and poor curves fall on or below the 45 degree line.

You can use this curve to determine a cutoff point for your predictive model, balancing sensitivity and specificity according to the costs of incorrect predictions.

See Receiver Operating Characteristics (ROC) Curves for more information.

•

ROC Statistics Plot

This overlay plot portrays several common performance statistics across all possible cutoffs of the posterior probability. Statistics include Specificity, Sensitivity, Accuracy, Positive Predictive Value, Negative Predictive Value, Matthews Correlation Coefficient, Youden Index, and Zero-One Index.

You can use these curves to determine a cutoff point for your predictive model, balancing the various criteria according to the costs of incorrect predictions.

See Receiver Operating Characteristics (ROC) Curves for more information.