An Essentially Complete Class of Two-Stage Selection Procedures with Screening at the First Stage.

Abstract

Let (pi sub 1),...,(pi sub k) be given populations associated with unknown real parameters (theta sub 1),...,(theta sub k). The goal is to find a population with a sufficiently large parameter in two stages with screening out inferior populations at the first stage. Both, the control and the non-control situations are considered simultaneously. Let sub I denote the class of permutation invariant randomized procedures (thi, psi, gamma), where at Stage 1, thi and psi decide how many populations and then which ones, respectively, are selected, and where at Stage 2, after additional samples have been drawn from the selected populations, gamma makes the final decision. Let psi and gamma denote the natural decisions, i.e. which are associated with the largest sufficient statistics. Under the assumption of a common discrete or continuous type strongly unimodal exponential family it is shown that with respect to every reasonable loss function, procedures of the type (thi, psi, gamma) form an essentially complete class within d sub I. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 01, 1981
Accession Number
ADA108065

Entities

People

  • Klaus-j. Miescke
  • Shanti Gupta

Organizations

  • Purdue University

Tags

Communities of Interest

  • Human Systems

DTIC Thesaurus Topics

  • Classification
  • Contracts
  • Convolution
  • Data Science
  • Decision Theory
  • Information Science
  • Military Research
  • New York
  • Observation
  • Permutations
  • Probability
  • Probability Distributions
  • Statistical Decision Theory
  • Statistics
  • Theorems
  • Universities

Fields of Study

  • Mathematics

Readers

  • Analytical Mechanics
  • Molecular and genetic basis of cancer.
  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference
  • AI & ML - Machine Learning Algorithms