On Bayesian interval prediction of future generalized-order statistics using doubly censoring

摘要

Based on a one-sample scheme, general Bayesian prediction intervals (BPI) for future generalized-order statistics are obtained when the previous and future samples are assumed to follow a general class of continuous distributions. The prior belief of the experimenter is measured by two distributions according to whether one (two) parameter(s) is (are) unknown. BPI for upper-order statistics and upper record values are obtained as special cases. Doubly Type II censored of the observed data has been used here. Application to the Weibull (θ_1, θ_2) model is illustrated when θ_1 is an unknown parameter and when bothθ_1 and θ_2 are unknown parameters. Numerical computations are made when H is unknown to illustrate the procedures.