Accuracy (tp + tn)/(number of observations).
The squared distance to the top left corner (1 − se)² + r(1 − sp)².
False negatives.
False positives.
False positive rate fp/(fp + tn).
Negative likelihood ratio (1 − se)/sp.
Positive likelihood ratio se/(1 − sp).
Negative predictive value tn/(tn + fn) (NaN when no observation is negative).
Positive predictive value tp/(tp + fp) (NaN when no observation is positive).
Sensitivity tp/(tp + fn).
Specificity tn/(tn + fp).
The threshold of each point.
True negatives.
True positives.
R's warnings, each once, in R's order (an empty array when there are none).
pROC's Youden criterion se + r·sp (not se + sp − 1).
Result of coords: one array per column, one value per point (pROC's data frame with
ret = "all").