Validation of a Monte Carlo Simulation of Binary Time Series.
Abstract
This report presents a statistical technique appropriate for the validation of a Monte Carlo simulation of a detection process whose results can be reduced to a finite time sequence of equally spaced events with dichotomous outcomes. The technique may be applied to other similar processes where there is not stationarity or steady state in the behavior of the process simulated. It is particularly useful in the case where multiple replications of the simulation have been obtained but only one observation of the real-world process is available. The Bahadur-Lazarsfeld representation of the probability distribution of the population consisting of all binary vectors having a given number of elements is generated from the sample of such vectors produced by several independent replications of the Monte Carlo simulation. Then the validity of the simulation is determined to within a specified significance level, by comparison with a single realization of the actual process simulated. This validation technique admits correlation between events occurring at different times. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 18, 1981
- Accession Number
- ADA104525
Entities
People
- D. Roque
- H. L. Wiener
- R. E. Johnson
Organizations
- United States Naval Research Laboratory