TESTING A SUBSET OF THE OVERIDENTIFYING RESTRICTIONS,

Abstract

Anderson and Rubin derived the likelihood ratio statistic for testing a particular overidentified model against the alternative that the model is just or underidentified. One problem with this test, as emphasized by Christ among others, is that consequences of rejecting the null hypothesis are not clear: what restrictions were wrong. Additionally, possibly some of the restrictions are suspect but others, enough to overidentify the model, are not. How the suspect overidentifying restrictions can be tested is the subject of the paper. (Author)

Document Details

Document Type
Technical Report
Publication Date
Jun 19, 1970
Accession Number
AD0708349

Entities

People

  • Joseph B. Kadane

Organizations

  • Center for Naval Analyses

Tags

Readers

  • Regression Analysis.