varoc (version 0.4.0)

varoc: VAROC: value added receiver operating characteristics (ROC) curve

Description

ROC curve to visualize classification and continuity performances of biomarkers, diagnostic tests, or risk prediction models.

Usage

varoc(y,x,tmd.range=NULL,legend="right",lwd=3,digits=2)

Value

th

Threshold values.

tpf

True positive fraction at each threshold.

fpf

False positive fraction at each th.

tpm

True positive mean at each th.

fpm

False positive fraction at each th.

tmd

Tail mean difference, i.e., tpm-fpm, at each th.

auc

Area under the ROC curve.

itmd

Integrated tmd over all theresholds.

Arguments

y

binary outcome, where y=1 if disease (or case) and y=0 if non-disease (or control).

x

continuous score, e.g. biomarker, diagnostic test, risk score.

tmd.range

minimum and maximum values of TMD, displayed on the plot.

legend

legend location, "bottomright", "bottom", "bottomleft", "left", "topleft", "top", "topright", "right" and "center".

lwd

line width.

digits

number of decimals.

Author

Yunro Chung [aut, cre]

Details

The varoc function summarizes a continuity performance of x at each cutoff c using two key metrics: (i) tail mean difference (TMD) and (ii) intergrated TMD (ITMD). For (i), TMD(c) is true positive mean(TPM)(c) minus false positive mean(FPM)(c), where TPM(c) is E(x>c|y=1) and FPM(c) is E(x>c|y=0). For (ii), ITMD is a global measure of evaluating continuity performance of x over all thresholds.

These measures are continuous versions of ROC curve-based measures. Specifically, TPM(c) and FPM(c) are continuous versions of true positive fraction (TPF)(c) and false positive fraction (FPF)(c), where TPF(c)=P(x>c|y=1) and FPF(c)=P(x>c|y=0). Thus, the useful (or useless) x has TPF(c)-FPF(c)>0 and TMD(c)>0 (or TPF(c)-FPF(c)=0 and TMD(c)=0); and useful (or useless) x has area under the ROC curve (AUC)>0.5 and ITMD(c)>0 (or AUC=0.5 and ITMD(c)=0).

References

Danielle Brister and Yunro Chung, Value added receiver operating characteristics curve (in-progress)

Examples

set.seed(100)

n1=100
n0=100
y=c(rep(1,n1),rep(0,n0))

#1. useless marker
x1=abs(c(rnorm(n1,0,1),rnorm(n0,0,1)))
fit1=varoc(y=y,x=x1)

#2. useful marker
x2=abs(c(rnorm(n1,2,1),rnorm(n0,0,1)))
fit2=varoc(y=y,x=x2)

#4. markers 1 vs 2
opar=par(mfrow=c(1,2))
tmd.range=range(c(fit1$tmd,fit2$tmd))
fit1=varoc(y=y,x=x1,tmd.range=tmd.range)
fit2=varoc(y=y,x=x2,tmd.range=tmd.range)
on.exit(par(opar))