Two ARMA-Based Confidence-Interval Procedures for the Analysis of Simulation Output.

Abstract

Two methods are presented for building interval estimates on the mean of a stationary stochastic process. Both methods fit an autoregressive moving-average (ARMA) model to observations on the process. The model is used to estimate the variance of the sample mean and the applicable degrees of freedom of the t statistic. Fitting of the ARMA model is totally automated. The ARMA-based confidence intervals perform well with data generated from ARMA processes. With data generated from queuing-system simulations, the coverage of the confidence intervals is less than satisfactory. It is shown that with queing-system data, sample mean and its estimated standard deviation are strongly positively correlated, and that the residuals of the fitted models are not normally distributed. These factors contribute adversely to the coverage of the confidence-interval procedures with queuing data. (Author)

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 01, 1982
Accession Number
ADA114934

Entities

People

  • Richard W. Andrews
  • Thomas J. Schriber

Organizations

  • University of Michigan

Tags

DTIC Thesaurus Topics

  • Business Administration
  • Computations
  • Data Science
  • Data Sets
  • Identification
  • Information Processing
  • Information Science
  • Materials Handling
  • Measures Of Effectiveness
  • New York
  • Normality
  • Plastic Explosives
  • Random Variables
  • Standards
  • Stationary Processes
  • Statistics
  • Stochastic Processes

Fields of Study

  • Mathematics

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Regression Analysis.