roc {caret}R Documentation

Compute the points for an ROC curve

Description

Computes sensitivity and specificity for a variety of cutoffs

Usage

roc(data, class, dataGrid = TRUE, gridLength = 100, positive = levels(class)[1])

Arguments

data a numeric variable to cut along
class a factor with class memberships. There must be only two classes.
dataGrid should the data define the grid of cut-points? If not a sequence of evenly spaced intervals is used.
gridLength number of intervals to use if the data do not define the grid.
positive a character string for the level of the class variable that defines a "positive" event

Value

A matrix of results with columns "cutoff", "sensitivity" and "specificity"

Note

The first row in the output has a cutoff of NA, zero sensitivity and specificity of one.

Author(s)

Max Kuhn

See Also

sensitivity, specificity, aucRoc

Examples

set.seed(6)
testData <- data.frame(
                       x = c(rnorm(200), rnorm(200) + 1),
                       group = factor(rep(letters[1:2], each = 200)))
   
densityplot(~testData$x, groups = testData$group, auto.key = TRUE)   
   
roc(testData$x, testData$group)

[Package caret version 4.10 Index]