Symmetric tests and confidence intervals for survival
probabilities and quantiles of censored survival data
Stuart Barber & Christopher Jennison.
We describe existing tests and introduce two new tests
concerning the value of a survival function. These tests may be used to
construct a confidence interval for the survival probability at a given
time or for a quantile of the survival distribution. Simulation studies
show that error rates can differ substantially from their nominal values,
particularly at survival probabilities close to 0 or 1. We recommend our
new 'constrained bootstrap' test for its good overall performance.
Some key words:
Beta distribution, bootstrap sampling, censoring, confidence interval,
Greenwood's formula, hypothesis test, Kaplan-Meier estimate,
likelihood ratio test, survival data.
Back to publications list