The Purpose of Mixed-Effects Models in Test and Evaluation

Abstract

All analyses of operational test data should account for uncertainty. Data from operational tests typically include statistical noise contributed from multiple sources. Mixed models, a standard technique for analyzing observations with an intra-group correlation structure, are a common way to model this type of data. The simplest version of the mixed model is the random intercept model, where the so-called random effects represent group-wide deviations from a grand mean. This approach can account for day-to-day deviations in system performance while still allowing the results to be generalized beyond the few days of observed testing. The R package ciTools allows analysts to generate uncertainty estimates (such as confidence intervals, prediction intervals, and quantile estimates) for mixed models (in addition to other, simpler types of linear models) quickly and efficiently. In some cases, users may wish to report uncertainty bounds that include the day-to-day variability, while in other cases, uncertainty bounds conditioned on particular days may be more relevant. ciTools facilitates both types of uncertainty estimation, allowing analysts to specify the uncertainty estimate they wish to use.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Aug 29, 2019
Accession Number
AD1122299

Entities

People

  • Heather Wojton
  • John T. Haman
  • Matthew R. Avery
  • Rebecca Medlin

Organizations

  • Institute for Defense Analyses

Tags

Communities of Interest

  • Air Platforms
  • Autonomy

DTIC Thesaurus Topics

  • Abstracts
  • Aircrafts
  • Algorithms
  • Analysis Of Variance
  • Contracts
  • Data Analysis
  • Data Science
  • Department Of Defense
  • Electronic Warfare
  • Governments
  • Information Operations
  • Information Science
  • Information Systems
  • Instructions
  • Intervals
  • Observation
  • Probability
  • Radar
  • Security
  • Standards
  • Test And Evaluation
  • Uncertainty
  • Unmanned Aerial Vehicles
  • Virginia

Readers

  • Computational Modeling and Simulation
  • Regression Analysis.