Receiver operating characteristic (ROC)

Share:

Description

To compute and draw a ROC curve.

Usage

1
  roc(y, x, z=NULL, event="less")

Arguments

y

Logical. Actual status: event or not.

x

Numberical vector based on which the diagnosis is performed.

z

Logical. A covariate for diagnosis in addition to 'x'.

event

Default: less. To classify a case to be an event if x is less than a cutoff. Otherwise, a case will be classified as an event of if x is greater than a cutoff.

Value

The TPR or sensitivity, and FPR or (1-specificity) will be returned together with the cutoffs.

The area under the ROC curve can be computed using summary function.

The ROC curve can be ploted using plot function.

Author(s)

B. Wang bwang@southalabama.edu

Examples

1
2
3
x <- rnorm(100)
y <- rnorm(100)<0
out = roc(y,x)