ABSTRACT
An initial step in the analysis of survival data is to present numerical or graphical summaries of observed survival times, based on the survivor function. The Kaplan-Meier and Nelson-Aalen estimates of the survivor function are considered in detail, and it is shown how standard errors of the estimated survivor functions can be found. Methods for estimating the median or other percentiles of the survival times are described, as well as the use of confidence intervals to summarise their precision. Estimation of the hazard and cumulative hazard functions is also summarised. Following a general introduction to hypothesis testing procedures, tests for comparing two groups of survival times are introduced. The log-rank, Wilcoxon and Peto-Peto tests are described in detail. It is also shown how these test procedures can be extended to enable three or more groups to be compared. The log-rank test for trend and test procedures for stratified survival data are also outlined.
