survcomp (version 1.22.0)

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

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

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

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.

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)