roc: Calculate Receiver Operating Characteristic (ROC) curve

In fmsb: Functions for Medical Statistics Book with some Demographic Data

Calculate Receiver Operating Characteristic (ROC) curve

Description

Calculate Receiver Operating Characteristic (ROC) curve's each performance set of [sensitivity, 1-specificity], each distance of the performance from the worst performance [0, 1], and each piece of area under the curve, for each cutoff point, as list. Fittest cut off is suggested as the set of [sensitivity, 1-specificity] which gives the longest distance from [0, 1] (though it's not common). If option maxdist=FALSE is given, the distances are calculated from the best performance [1, 0] and fittest cut off is the set of [sensitivity, 1-specificity] which gives minimum distance from best performance.

Usage

 roc(values, iscase, maxdist=TRUE) 

Arguments

values	A numeric vector of measured values.
iscase	A logical (or 0/1) vector of diagnostics. Negative result must be given by FALSE or 0.
maxdist	A logical value to specify the method of distance calculation to seek the best cutoff. Default TRUE.

Value

cutoff	The numeric vector of cutoff points, which are composed of the all unique values among the given measurements and the maximum cutoff is maximum measurement plus 1. Therefore, the minimum cutoff gives [1, 1] and the maximum cutoff gives [0, 0] as the performance set of [sensitivity, 1-specificity], respectively.
sens	The numeric vector of sensitivities for all cutoff points.
falsepos	The numeric vector of 1-specificities (false positiveness) for all cutoff points.
distance	The numeric vector of distance between actual performance set and the worst performance.
aucpiece	The numeric vector of the pieces of areas under the curve.
maxdist	Same as the given argument maxdist.

Author(s)

Minato Nakazawa [e-mail] https://minato.sip21c.org/

See Also

ROC

Examples

 scores <- c(15, 20, 19, 28, 26, 17, 13, 22, 23, 24)
 diagno <- c(0, 0, 0, 1, 1, 1, 0, 1, 1, 1)
 res <- roc(scores, diagno)
 print(res)
 plot(res)

fmsb documentation built on Jan. 6, 2023, 1:18 a.m.