Simultaneous Confidence Intervals For Differences Of Means From Several Two-parameter Exponential Distributions

In this thesis, simultaneous confidence intervals are constructed for all pairwise differences of means from several two-parameter exponential distributions.Firstly, this thesis propose fiducial generalized confidence intervals method to construct simultaneous confidence intervals for all pairwise differences of means from two-parameter expo-nential distributions. The proposed method is based on the concept of the Fiducial Generalized Pivotal Quantity and it is shown that the confidence intervals have correct coverage probability asymptotically. Finite sample performance of the proposed simultaneous confidence intervals is evaluated by simulation studies, and the results indicate that the empirical coverage probabilities attains the nominal confidence level.Secondly, a parametric bootstrap method is proposed for constructing simultaneous con-fidence intervals for all pairwise differences of means from several two-parameter exponential distributions. The proposed simultaneous confidence intervals are shown to have correct cov-erage probability asymptotically. Simulation studies show that, comparing with some existing methods, the parametric bootstrap simultaneous confidence intervals are generally closer to the nominal level and possess smaller volumes.