Estimation in Parametric Mixture Families.
Abstract
For parametric mixture distributions indexed by say a univariate parameter Theta we investigate estimation of g theta under squared error loss. First we propose a method for uniformly improving upon an unbaised estimator of g theta. Second we characterize Bayes estimators of g theta and give a simple complete class theorem. Finally we study the performance of empirical Bayes rules generated using the EM algorithm. Application in the context of the noncentral chi-square distribution provides examples.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 18, 1987
- Accession Number
- ADA183818
Entities
People
- Alan E. Gelfand
Organizations
- Stanford University