Optimization in Analytical Chemistry Using Robust Estimation.

Abstract

Analytical chemists have long been concerned with obtaining optimal experimental conditions. robust estimation provides an additional method of increasing the efficiency of an analytical technique. This is illustrated for the determination of the true value, u, of a quantity which is measured with error. The least squares estimator of u is compared with the median and Huber estimates over a variety of error distributions in the vicinity of the Gaussian distribution. Simulation allows examination of the efficiency of an estimation procedure as a function of the error distribution. Results are presented which show the least squares estimator of u to be much more sensitive to a non-Gaussian error distribution than generally realized in the chemical community. Additionally, the arguments commonly used to support least squares estimation are critically examined. Keywords: Robust estimation, least squares estimator, Huber estimator, Gaussian distribution, non Gaussian error distribution.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 30, 1987
Accession Number
ADA183326

Entities

People

  • Edward M. Eyring
  • Gregory R. Phillips

Organizations

  • University of Utah

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Analytical Chemistry
  • Artificial Intelligence
  • Chemical Analysis
  • Chemistry
  • Classification
  • Computers
  • Contamination
  • Data Analysis
  • Estimators
  • Gaussian Distributions
  • Information Science
  • Notation
  • Optimal Estimators
  • Probability
  • Random Variables
  • Statistics
  • Universities

Readers

  • Statistical inference.
  • Theoretical Analysis.