Files in this item
|(no description provided)|
|Title:||Admissibility ofp-value rules|
|Author(s):||Thompson, Peter Michael|
|Doctoral Committee Chair(s):||Marden, John I.|
|Department / Program:||Statistics|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Abstract:||This work deals with a decision-theoretic evaluation of p-value rules. A test statistic is judged on the behavior of its p-value with the loss function being an increasing function G of the p-value.
Examples are presented in which every fixed-level test based on a statistic is admissible, but the corresponding p-value is inadmissible. It is also shown that if a statistic has an admissible p-value, all of its corresponding fixed-level tests are admissible.
In addition, Bayes p-values are investigated. It is shown that Bayes p-values and their weak limits form an essentially complete class. Also, the set of p-values which are Bayes typically does not depend on the particular loss function G being used.
Finally, the case where the parameter space is compact is looked into. A characterization is given of an essentially complete class of p-values.
|Rights Information:||Copyright 1989 Thompson, Peter Michael|
|Date Available in IDEALS:||2011-05-07|
|Identifier in Online Catalog:||AAI9011052|