Optimal equivariant prediction regions based on multiply type-Ⅱ censored generalized order statistics from exponential distributions

摘要

Interval prediction of future generalized order statistics from exponential distributions based on a multiply Type-II censored sample is considered. On invoking the principle of reduction by equivariance, prediction regions of minimal size are derived and shown to be of interval form. An alternative derivation based on the use of a prediction sufficient statistic is pointed out.