Efficient Identification of Important Factors in Large Scale Simulations

Abstract

Large, complex computer simulation models can require prohibitivelly costly and time-consuming experimental programs to study their behavior. Therefore we may want to concentrate the analysis on the set of most important factors (i.e., input variables). Factor screening experiments, which attempt to identify the more important variables, cana be extremely useful in the study of such models. The number of computer runs available for screening, however, is usually severely limited. In fact, the number of factors often exceeds the number of available runs. This paper presents a survey of supersaturated designs for use in factor screening experiments. The designs considered are: random balance, systematic supersaturated, group screening, modified group screening, T-optimal, R-optimal, and search designs. Discussed in terms are the basic technique, advantages, and disadvantages of each procedure surveyed.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Oct 01, 1986
Accession Number
ADA173497

Entities

People

  • Carl A. Mauro

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Combinatorial Analysis
  • Computational Science
  • Computer Simulations
  • Computers
  • Data Science
  • Design Criteria
  • Experimental Design
  • Factorial Design
  • Information Science
  • New York
  • Operations Research
  • Probability
  • Regression Analysis
  • Simulations
  • Statistics
  • Surveys
  • United States

Readers

  • Computational Modeling and Simulation
  • Systems Analysis and Design