VV&A of M&S in the DoD: Moving from Art to Science

Abstract

Many a statistician or serious student of statistics will recall his or her introduction to the great George Box quote: "All models are wrong; some are useful." Some of us first saw it in the response surfaces text written by Box and Draper (1987), but the quote is actually attributed to Box in one of his much earlier texts (Box 1979; Box and Draper 1987). We know from Dr. Box's good teachings that to improve the credibility of using a regression model the modeler should prove a number of assumptions (e.g., independence in observations, distribution of residuals is normal around 0) and should test that the model does not significantly overfit or underfit the data (i.e., lack-of-fit tests). While the exact tests used for these measures do not precisely port to the validation of modeling and simulation (M&S), the goal of reporting the observation is a healthy one: Treat the validation of Department of Defense (DoD) M&S more as a science and less as an art.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 2010
Accession Number
ADA530390

Entities

People

  • Amy Henninger
  • John Diem
  • Simone Youngblood

Tags

Communities of Interest

  • Autonomy

DTIC Thesaurus Topics

  • Acquisition
  • Artificial Intelligence
  • Computer Science
  • Department Of Defense
  • Distributed Interactive Simulations
  • Electronic Mail
  • Engineering
  • Machine Learning
  • Multiagent Systems
  • Neural Networks
  • Simulations
  • Standards
  • Statistics
  • Students
  • Systems Engineering
  • Test And Evaluation
  • United States

Readers

  • Computational Modeling and Simulation
  • Military History of the United States in the 20th Century.
  • Regression Analysis.