On Bayes and Gamma-Minimax Subset Selection Rules.

Abstract

Selection and ranking problems arise because the classical tests of homogeneity are often inadequate in practice when the experimenter wishes to make decisions regarding k(> or = 2) populations or treatments. Chapter 1 deals with the problem of selecting 'good' populations from a set of k given populations pi sub 1,...,pi sub k where the quality of each population pi sub i is characterized by an unknown parameter theta sub i. A selection problem arising in reliability theory and another for the selection in terms of scale parameters are considered. The first problem deals with the selection of components (units) for parallel and series systems from k populations (brands) with exponentially distributed life-length times. An optimal rule is derived for the series system and the Bayes rule with respect to a natural conjugate prior is derived for the 1-out-of-2 system. The second problem deals with the investigation of the selection procedures based on robust estimators of measures of dispersion for selecting the populations in terms of scale parameters. Large sample solutions and the asymptotic relative efficiencies of the proposed procedures are studied.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 1979
Accession Number
ADA069014

Entities

People

  • Woo-chul Kim

Organizations

  • Purdue University

Tags

Communities of Interest

  • C4I

DTIC Thesaurus Topics

  • Applied Mathematics
  • Bayesian Networks
  • Computational Science
  • Computations
  • Data Science
  • Distribution Functions
  • Estimators
  • Information Science
  • New York
  • Normal Distribution
  • Probability
  • Probability Density Functions
  • Random Variables
  • Statistical Algorithms
  • Statistical Inference
  • Statistics
  • Theorems

Fields of Study

  • Mathematics

Readers

  • Regression Analysis.
  • Statistical inference.