Development of two-stage methods for non-normal populations

提出了两种构造非正态总体均值向量置信区域的新方法。一种是利用分布渐近展开的方法，另一种是利用自举法的方法。我们阐明了每种方法的特点，包括它们的缺点。我们发现了另一种自举方法来克服原始Healy两阶段方法中样本量估计过高的问题

Two new methods of two-stage procedure of constructing a confidence region of the mean vector for non-normal populations are suggested. The one is a method using asymptotic expansion of the distribution, and the other is a method using a bootstrap method. We clarified character of each methods including their drawback. We found another bootstrap method to overcome a problem of over estimating the sample size needed in the original Healy' s two-stage method