CONFIDENCE INTERVAL OF SURVIVAL FUNCTION FOR EXTREME VALUE DISTRIBUTION WITH CENSORED DATA

Extreme value distribution has wide applicability in the parametric analysis of lifetime data. For the extreme value distribution, we construct confidence intervals for the survival function with censored data. The bootstrap method and its logistic transformation are the basis of the confidence intervals' construction. For comparison, we also consider the following confidence intervals: the normal approximation, the normal approximation with log and arcsine transformations and the likelihood ratio (LR) method. Finite sample behaviour of the six confidence intervals is examined through simulation. The confidence intervals are applied to a real world data set.