confIntKM: Confidence intervals for a survival curve at a fixed time...

View source: R/confIntKM.R

confIntKMR Documentation

Confidence intervals for a survival curve at a fixed time point

Description

Computes Wald confidence intervals for a Kaplan-Meier survival curve at a fixed time point. The variance is computed according to Peto's formula and the confidence interval is computed using a logit-transformation, to ensure that it lies in (0, 1). Alternatives are given in the examples below.

Usage

confIntKM(time, event, t0, conf.level = 0.95)

Arguments

time

Event times, censored or observed.

event

Censoring indicator, 1 for event, 0 for censored.

t0

Numeric vector of time points to compute confidence interval for.

conf.level

Confidence level for confidence interval.

Value

A matrix with the following columns:

t0

Time points.

S_t0

Value of survival curve at t0.

lower.ci

Lower limits of confidence interval(s).

upper.ci

Upper limits of confidence interval(s).

Note

The variance according to Peto's formula tends to be more conservative than that based on Greenwood's formula.

Author(s)

Kaspar Rufibach
kaspar.rufibach@gmail.com

Examples


## use Acute Myelogenous Leukemia survival data contained in package 'survival'
time <- leukemia[, 1]
event <- leukemia[, 2]
formula <- Surv(time = time, event = event) ~ 1
plot(survfit(formula = formula, conf.type = "none"), mark = "/", col = 1:2)
confIntKM(time = time, event = event, t0 = c(10, 25, 50), conf.level = 0.95)


## an alternative is the log-log confidence interval using Greenwood's
## variance estimate
t0 <- 10
fit <- survfit(formula = formula, conf.int = 0.95, conf.type = "log-log", 
               type = "kaplan", error = "greenwood")
dat <- cbind(fit$time, fit$surv, fit$lower, fit$upper)
dat <- dat[dat[, 1] >= t0, ]
dat[1, 3:4]


## this same confidence interval can also be computed using the 
## package km.ci
if(require(km.ci)) {
    ci.km <- km.ci(survfit(formula = formula), conf.level = 0.95,
                   method = "loglog")
    dat.km <- cbind(ci.km$time, ci.km$surv, ci.km$lower, ci.km$upper)
    dat.km <- dat.km[dat.km[, 1] >= t0, 3:4]
    dat.km[1, ]
}


felix-hof/biostatUZH documentation built on Sept. 27, 2024, 1:48 p.m.