Area under the ROC curve

The area under the ROC curve is a good measure of predictive accuracy for a binary event. Given a dataset with observed (0/1) and predicted values (probabilities), the function roc will calculate for a series of cutpoints the sensitivity, specificity, likelihood ratios and the area under the ROC curve. This function will also produce three plots.

data(Ionosphere)
model <- glm(Class ~ ., data=Ionosphere[,15:35], family=binomial)
roc(predict(model,type="response"), Ionosphere$Class=="good")

cp sens espe
[1,] 0.00 1.00000000 0.0000000
[2,] 0.05 1.00000000 0.2380952
[3,] 0.10 0.99555556 0.3412698
[4,] 0.15 0.99555556 0.3809524
[5,] 0.20 0.99555556 0.4206349
[6,] 0.25 0.99111111 0.4365079
[7,] 0.30 0.99111111 0.4603175
[8,] 0.35 0.98222222 0.5000000
[9,] 0.40 0.97777778 0.5317460
[10,] 0.45 0.97333333 0.5634921
[11,] 0.50 0.96444444 0.6111111
[12,] 0.55 0.94222222 0.6507937
[13,] 0.60 0.92000000 0.7301587
[14,] 0.65 0.85333333 0.7777778
[15,] 0.70 0.77333333 0.7936508
[16,] 0.75 0.69333333 0.8174603
[17,] 0.80 0.59555556 0.8412698
[18,] 0.85 0.40888889 0.8888889
[19,] 0.90 0.20000000 0.9285714
[20,] 0.95 0.01777778 0.9603175
[21,] 1.00 0.00000000 1.0000000