Experimental Designs for Quantal Response Models.

Abstract

Three problems have been studied. A new sequential design procedure has been developed for estimating the percentiles of a quantal response curve which describes the probability of response as a function of stimulus level. It is asymptotically fully efficient and distribution-free. For small samples it outperforms the best Robbins-Monro procedure in an extensive empirical study. The percentages of saving range from 25% to 60%. It is shown that, by classifying the outcome of a sensitivity experiment into more than two categories, the parameters in a parametric response curve can be estimated more precisely. The theoretical result is obtained via the Missing Information Principle. The empirical study points out where substantial gains can be achieved. Several procedures for the multiple comparison of all linear contrasts in unbalanced situations are compared both empirically and theoretically. Recommendations for the choice of procedures are given. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Sep 22, 1983
Accession Number
ADA133416

Entities

People

  • Chien-fu Jeff Wu

Organizations

  • University of Wisconsin Madison Department of Statistics

Tags

DTIC Thesaurus Topics

  • Analysis Of Variance
  • Classification
  • Contrast
  • Data Science
  • Experimental Design
  • Information Science
  • Military Research
  • Probability
  • Security
  • Sensitivity
  • Simulations
  • Statistics
  • Universities
  • Wisconsin

Fields of Study

  • Mathematics

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Regression Analysis.