Abstract
Exact confidence intervals are currently obtainable for parameters, percentiles, and reliability in the two-parameter Weibull (or extreme-value) model with complete or type II censored data, but not when the data are progressively censored. This article proposes a simple approximation for this last case. The algorithm for computing confidence limits is a type of bootstrapping. The method is applicable to any model that is transformable to location-scale form and has invariant estimators. Both fixed and random censoring are discussed. A numerical example and some simulation results are given for Weibull confidence intervals based on maximum likelihood estimators.