| bootkm {Hmisc} | R Documentation |
Bootstraps Kaplan-Meier estimate of the probability of survival to at
least a fixed time (times variable) or the estimate of the q
quantile of the survival distribution (e.g., median survival time, the
default).
bootkm(S, q=0.5, B=500, times, pr=TRUE)
S |
a Surv object for possibly right-censored survival time
|
q |
quantile of survival time, default is 0.5 for median |
B |
number of bootstrap repetitions (default=500) |
times |
time vector (currently only a scalar is allowed) at which to compute
survival estimates. You may specify only one of q and times, and
if times is specified q is ignored.
|
pr |
set to FALSE to suppress printing the iteration number every 10 iterations
|
bootkm uses Therneau's survfit.km function to efficiently compute
Kaplan-Meier estimates.
a vector containing B bootstrap estimates
updates .Random.seed, and, if pr=TRUE, prints progress of simulations
Frank Harrell
Department of Biostatistics
Vanderbilt University School of Medicine
f.harrell@vanderbilt.edu
Akritas MG (1986): Bootstrapping the Kaplan-Meier estimator. JASA 81:1032–1038.
survfit, survfit.km, Surv, Survival.cph, Quantile.cph
# Compute 0.95 nonparametric confidence interval for the difference in
# median survival time between females and males (two-sample problem)
set.seed(1)
library(survival)
S <- Surv(runif(200)) # no censoring
sex <- c(rep('female',100),rep('male',100))
med.female <- bootkm(S[sex=='female',], B=100) # normally B=500
med.male <- bootkm(S[sex=='male',], B=100)
describe(med.female-med.male)
quantile(med.female-med.male, c(.025,.975), na.rm=TRUE)
# na.rm needed because some bootstrap estimates of median survival
# time may be missing when a bootstrap sample did not include the
# longer survival times