Confidence Regions Based on Methods of Combining Test Statistics
本文研究了将多种检验统计量合并方法反演，得到参数向量的联合置信域，并讨论了这些区域的性质，通过实例说明其应用。
Abstract A basic problem in statistics concerns the combination of test statistics. The most widely studied combination procedure is attributed to R.A. Fisher (1950) and is based on the product of the observed significance levels of the individual tests. Other methods include the likelihood ratio and a procedure based on the minimum significance level. In this article we consider inverting test combination methods to obtain simultaneous confidence regions for a parameter vector θ. Properties of the regions thus obtained are discussed in terms of known results in the corresponding test combination framework. Several examples are presented to illustrate the procedures.