Confidence Intervals for the Common Variance of Equicorrelated Normal Random Variables.
Abstract
Confidence intervals for the common variance of equicorrelated normal random variable are investigated. When the correlation coefficient is unknown no uniformly most accurate invariant confidence interval for the common variance exists. We develop interval estimators which are based on the asymptotically most efficient ones by employing the best asymptotic normality of the maximum likelihood estimators. In addition to their comparison with possible competitors, their small sample efficiency is studied. A method for determining exact coverage probabilities and expected length is developed. As a result of these investigations we concluded that interval estimators based on maximum likelihood estimators are to be recommended.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1979
- Accession Number
- ADA074339
Entities
People
- P. F. Ramig
- Shelemyahu Zacks
Organizations
- Case Western Reserve University