Empirical Bayes Estimation of Critical Dosages Having Smallest Predictive Risk.
Abstract
An empirical Bayes procedure is developed for the estimation of critical dosages in the linear regression case. If Y = alpha + beta x + e is the basic linear model, the critical dosage is defined as xi (eta) = (eta - alpha)/beta, for beta > 0. A new type of Bayes estimator of xi(eta) is derived under the criterion of minimizing the predictive risk E < (alpha + beta xi - eta)-squared bar F sub n >. The empirical Bayes procedure provides consistant estimators of the prior parameters when a large number of independent repetitions of the experiment is available. The methodology is developed to analyze a large set of photodynamic bioassays, for the determination of critical air concentrations of benzo-soluble organic extracts. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 15, 1979
- Accession Number
- ADA074376
Entities
People
- Ora Bialik
- Shelemyahu Zacks
Organizations
- Case Western Reserve University