I have a concern about the survAUC package option AUC.cd.
I am exploring package functionality, specifically AUC statistics for Cox
Regression, for a small academic project
When utilizing this package on the ovarian data set within that package I
obtain an AUC statistic of 0.3322928. When AUC calculations use a
dichotomous outcome such as this, see included R code, the result should lie
between 0.5 and 1, not 0.33.
Please explain this, I am not certain that the algorithm that is being
utilized for this package is correct.
Thank you
Kevin O?Donnell, MS Work Environment, MS Env Eng., MS Const Project Mgmt
Graduate Student
Department of Biostatistics
Boston University School of Public Health
715 Albany Street Boston, MA
617-480-1677
x11(h=8,w=11)
fit = survfit(Surv(futime,fustat) ~ rx)
plot(fit, mark.time=FALSE, xscale=365.25,main="Plot of Survival Curves by
Prescription Status",
xlab='Length of Survival', ylab='Proportion of Individuals who have
Survived')
lines(fit[1], lwd=3,lty=2:3, xscale=365.24,col=2)
lines(fit[2], lwd=2,lty=2:2, xscale=365.24,col=3)
legend(.2,.2, c("No treatment", "treatment"), lwd=3,lty = 2:3)
TR2 = ovarian[1:16,]
TE2 = ovarian[17:26,]
train.fit2 = coxph(Surv(futime, fustat) ~ rx,
x=TRUE, y=TRUE, method="efron", data=TR)
lp2 = predict(train.fit)
lpnew2 = predict(train.fit2, newdata=TE2)
Surv.rsp2 = Surv(TR2$futime, TR2$fustat)
Surv.rsp.new2 = Surv(TE2$futime, TE2$fustat)
times2 = seq(10, 1000, 10)
AUC_CD2 = AUC.cd(Surv.rsp2, Surv.rsp.new2, lp2, lpnew2, times2)
AUC_hc2 = AUC.hc(Surv.rsp2, Surv.rsp.new2, lpnew2, times2)
AUC_sh2 = AUC.sh(Surv.rsp2, Surv.rsp.new2, lp2, lpnew2, times2)
AUC_Uno2 = AUC.uno(Surv.rsp2, Surv.rsp.new2, lpnew2, times2)
--
View this message in context: http://r.789695.n4.nabble.com/R-survAUC-Package-tp4681638.html
Sent from the R help mailing list archive at Nabble.com.
R survAUC Package
2 messages · kevinod, David Winsemius
2 days later
On Dec 4, 2013, at 7:49 AM, kevinod wrote:
I have a concern about the survAUC package option AUC.cd.
So shouldn't you be sending this to the package authors? They may or may not be regular readers of R help. It's apackage I have never heard of.
I am exploring package functionality, specifically AUC statistics for Cox
Regression, for a small academic project
When utilizing this package on the ovarian data set within that package I
obtain an AUC statistic of 0.3322928. When AUC calculations use a
dichotomous outcome such as this, see included R code, the result should lie
between 0.5 and 1, not 0.33.
Please explain this, I am not certain that the algorithm that is being
utilized for this package is correct.
Thank you
Kevin O?Donnell, MS Work Environment, MS Env Eng., MS Const Project Mgmt
Graduate Student
Department of Biostatistics
Boston University School of Public Health
715 Albany Street Boston, MA
617-480-1677
x11(h=8,w=11)
fit = survfit(Surv(futime,fustat) ~ rx)
plot(fit, mark.time=FALSE, xscale=365.25,main="Plot of Survival Curves by
Prescription Status",
xlab='Length of Survival', ylab='Proportion of Individuals who have
Survived')
lines(fit[1], lwd=3,lty=2:3, xscale=365.24,col=2)
lines(fit[2], lwd=2,lty=2:2, xscale=365.24,col=3)
legend(.2,.2, c("No treatment", "treatment"), lwd=3,lty = 2:3)
TR2 = ovarian[1:16,]
TE2 = ovarian[17:26,]
train.fit2 = coxph(Surv(futime, fustat) ~ rx,
x=TRUE, y=TRUE, method="efron", data=TR)
lp2 = predict(train.fit)
lpnew2 = predict(train.fit2, newdata=TE2)
Surv.rsp2 = Surv(TR2$futime, TR2$fustat)
Surv.rsp.new2 = Surv(TE2$futime, TE2$fustat)
times2 = seq(10, 1000, 10)
AUC_CD2 = AUC.cd(Surv.rsp2, Surv.rsp.new2, lp2, lpnew2, times2)
AUC_hc2 = AUC.hc(Surv.rsp2, Surv.rsp.new2, lpnew2, times2)
AUC_sh2 = AUC.sh(Surv.rsp2, Surv.rsp.new2, lp2, lpnew2, times2)
AUC_Uno2 = AUC.uno(Surv.rsp2, Surv.rsp.new2, lpnew2, times2)
--
View this message in context: http://r.789695.n4.nabble.com/R-survAUC-Package-tp4681638.html
Sent from the R help mailing list archive at Nabble.com.
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
David Winsemius Alameda, CA, USA