Technique for Increasing Efficiency and Accuracy of Data for Mix Analysis

Abstract

This report is an examination of the issues and possible solutions to the problem of statistical overlapping groups. The research which was the basis for the development of improved accuracy for a mix model analysis, was conducted jointly by the Training and Doctrine Command (TRADOC) Analysis Center-White Sands Missile Range (TRAC-WSMR) and New Mexico State University (NMSU) as a dissertation research project by the author. The objective of this research and the conclusion reached in this report is to establish a statistic based method for separating treatment means into distinct groups without group overlap, thereby increasing accuracy of data used for mix analysis. Additionally, using computer processing improves the efficiency of data preparation and adds improved accuracy derived from the developed method. Six different representative data sets were processed using five different multiple comparison procedures and heuristic algorithms. The results were analyzed and the procedures that were most effective at producing the best groupings were identified. The Base Grouping Heuristic provided the best results when used with Tukey's multiple comparison procedure with a significance level of 0.05 or with Fisher's Least Significant Difference procedure with a significance level of 0.01. The Base Grouping Heuristic was then programmed into an automated procedure.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 01, 1998
Accession Number
ADA350724

Entities

People

  • Bruce W. Gafner

Tags

Communities of Interest

  • Weapons Technologies

DTIC Thesaurus Topics

  • Accuracy
  • Algorithms
  • Computer Programs
  • Computers
  • Data Science
  • Data Sets
  • Databases
  • Doctrine
  • Information Processing
  • Information Science
  • New Mexico
  • Social Sciences
  • Statistical Analysis
  • Statistical Inference
  • Statistics
  • Training
  • United States

Readers

  • Military Training and Readiness Simulation
  • Operations Research
  • Regression Analysis.