Likelihood ratio statistics based on an integrated likelihood

• 成果类型: Article
• 署名作者: Severini, T. A.
• 署名单位: Northwestern University
• 刊物名称: BIOMETRIKA
• ISSN/ISSBN: 0006-3444
• DOI: 10.1093/biomet/asq015
• 发表日期: 2010
• 页码: 481496
• 关键词: profile likelihood; nuisance parameters; models; inference; priors; tests; FORMULA; POWER

摘要

An integrated likelihood depends only on the parameter of interest and the data, so it can be used as a standard likelihood function for likelihood-based inference. In this paper, the higher-order asymptotic properties of the signed integrated likelihood ratio statistic for a scalar parameter of interest are considered. These results are used to construct a modified integrated likelihood ratio statistic and to suggest a class of prior densities to use in forming the integrated likelihood. The properties of the integrated likelihood ratio statistic are compared to those of the standard likelihood ratio statistic. Several examples show that the integrated likelihood ratio statistic can be a useful alternative to the standard likelihood ratio statistic.