hanley {gcl}R Documentation

Statistically compare C-indices

Description

Compares two binary classifier outputs on the same data using the method by Hanley and McNeil.

Usage

hanley(paref1, paref2, aaref, ctype = "pearson")

Arguments

paref1 vector of predictions from classifier 1
paref2 vector of predictions from classifier 2
aaref actual classes, must be 0 or 1
ctype type of correlation used in cor.test

Details

Implements the method of Hanley and McNeil for comparing two c-indices (ROC AUC's) from the same data.

Value

Returns a list.

p p value
auc1 the AUC of classifier 1
auc2 the AUC of classifier 2
se1 standard error of AUC of classifier 1
se2 standard error of AUC of classifier 2
z z value
r Correllation factor
corr0s correlation of true 0's
corr1s correlation of true 1's

Author(s)

Staal A. Vinterbo staal@dsg.harvard.edu

References

J.A. Hanley and B.J. McNeil. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology, 148:839-43, 1983.

See Also

cindex

Examples

p1 <- runif(10)
p2 <- runif(10)
a <- round(runif(10))
hanley(p1,p2,a)

[Package gcl version 1.06.5 Index]