iauc.comp: Function to compare two IAUCs through time-dependent ROC...

In bhklab/survcomp: Performance Assessment and Comparison for Survival Analysis

Description

This function compares two integrated areas under the curves (IAUC) through the results of time-dependent ROC curves at some points in time. The statistical test is a Wilcoxon rank sum test for dependent samples.

Usage

iauc.comp(auc1, auc2, time)

Arguments

auc1	vector of AUCs computed from the first time-dependent ROC curves for some points in time
auc2	vector of AUCs computed from the second time-dependent ROC curves for some points in time
time	vector of points in time for which the AUCs are computed

Details

The two vectors of AUCs must be computed from the same samples (and corresponding survival data) and for the same points in time. The function uses a Wilcoxon rank sum test for dependent samples.

Value

p.value	p-value from the Wilcoxon rank sum test for the comparison iauc1 > iauc2
iauc1	value of the IAUC for the first set of time-depdent ROC curves
iauc2	value of the IAUC for the second set of time-depdent ROC curves

Author(s)

Benjamin Haibe-Kains

References

Wilcoxon, F. (1945) "Individual comparisons by ranking methods", Biometrics Bulletin, 1, pages 80–83.

Haibe-Kains, B. and Desmedt, C. and Sotiriou, C. and Bontempi, G. (2008) "A comparative study of survival models for breast cancer prognostication based on microarray data: does a single gene beat them all?", Bioinformatics, 24, 19, pages 2200–2208.

See Also

tdrocc, wilcox.test

Examples

set.seed(12345)
age <- rnorm(30, 50, 10)
size <- rexp(30,1)
stime <- rexp(30)
cens <- runif(30,.5,2)
sevent <- as.numeric(stime <= cens)
stime <- pmin(stime, cens)
##time-dependent ROC curves
tt <- unique(sort(stime[sevent == 1]))
##size
mytdroc1 <- NULL
for(i in 1:length(tt)) {
	rr <- tdrocc(x=size, surv.time=stime, surv.event=sevent, time=tt[i],
    na.rm=TRUE, verbose=FALSE)
	mytdroc1 <- c(mytdroc1, list(rr))
}
auc1 <- unlist(lapply(mytdroc1, function(x) { return(x$AUC) }))
##age
mytdroc2 <- NULL
for(i in 1:length(tt)) {
	rr <- tdrocc(x=age, surv.time=stime, surv.event=sevent, time=tt[i],
    na.rm=TRUE, verbose=FALSE)
	mytdroc2 <- c(mytdroc2, list(rr))
}
auc2 <- unlist(lapply(mytdroc2, function(x) { return(x$AUC) }))
iauc.comp(auc1=auc1, auc2=auc2, time=tt)

bhklab/survcomp documentation built on Dec. 26, 2021, 6:41 a.m.