The effectiveness of variance reduction techniques that users can apply to improve the efficiency and reliability of simulation experiments with the TRAF-NETSIM simulation model is described and illustrated. The two variance reduction techniques, antithetic variates and common random numbers, reduce the variance of simulation output by replacing the original sampling procedure by a new procedure that yields the same parameter estimate but with a smaller variance. Thus, the users can obtain greater statistical accuracy for the same number of simulation runs. A recent modification of the stochastic sampling process has made the TRAF-NETSIM model amenable to these variance reduction techniques and allows the users to apply these techniques with minimal additional effort. The effectiveness of these techniques is evaluated through an analysis of simulation output data from a TRAF-NETSIM case study. The estimated values and variances are computed for some representative measures of effectiveness after 10, 20, and 30 replications. The results indicate that both techniques are effective in reducing variance of the model output. By using the variance reduction techniques, the variance of parameter estimates is reduced on the average by 65% in the 24 comparisons that are made. The common random numbers strategy is more effective than the antithetic variates procedure. Over 50% reduction in variance is obtained using the common random numbers strategy in all comparisons and 80% or more in 6 of the 12 comparisons. In all cases studied, better statistical precision is obtained by making the two-thirds fewer simulations than under conventional multiple replications-based experimentation.


  • English

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 133-146
  • Monograph Title: Highway capacity and traffic flow
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00626853
  • Record Type: Publication
  • ISBN: 0309054044
  • Files: TRIS, TRB
  • Created Date: Feb 17 1993 12:00AM