The Comparison Of Some Simultaneous Confidence Intervals

In order to find simultaneous confidence intervals for finite or all the estimable functions in a normal linear model, the literature[1-3] gives us three methods:Scheffe, Bonfenoni and Tukey simultaneous confidence intervals. Thereinto, Tukeyinterval is mainly used in a balance one-way classification model ,which is a special case of a normal linear model. Popularly, given a normal linear model and some estimable functions ,the three methods are usable. Which one is the best? So, we should compare them with eath other.This thesis mainly investigate two topics: one side is the comparison of the three intervals in a fixed-size sampling. The other side is the comparison in two-stage sampling, namely.non-fixed-size sampling.In chapter 0, the author gives the introduction and some theorems,which are the base of this paper .And they have the important meaning to the author's work.In chapter 1, the author firstly units the three intervals into groups to discuss with the difference of models. The method of the discussion is to compare the three intervals' length (viz.precision) on the assumption that the converage probability (viz.reliability) is the same . The length is shorter, the interval is better. Here, using the mathematica software and the outcome of the computation , we can see which interval is better easily.In chapter 2, the author firstly studies three two-stage simultaneous confidence intervals for the pairwise treatments with the same precision and reliability in a balance one-way classification model ,then work out the optimal sample size of the first stage by numerial computation.By comparing the optimal sample size of the first stage and the average sample size, we can see which one is the best...