On the Bounded Regret of Empirical Bayes Estimators.
Abstract
In the first Jerzy Neyman Memorial Lecture, Robbins (1983) has outlined a wide class of problems concerning the general empirical Bayes approach and the linear empirical Bayes approach to estimation. This paper studies a special case which includes several important standard distributions. Specifically let (theta, x) be a random vector such that theta has a distribution function G, and the conditional expectation of x given theta satisfies E(x/theta) = theta. Suppose it is desired to use a linear function A+Bx of the observed x to estimate the unknown parameter theta.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1985
- Accession Number
- ADA169108
Entities
People
- Kai F. Yu
Organizations
- University of South Carolina