An Investigation of Finite Sample Behavior of Confidence Interval Estimation Procedures in Computer Simulation
Abstract
Investigated are the small sample behavior and convergence properties of confidence interval estimators (CIE's) for the mean of a stationary discrete process. We consider CIE's arising from nonoverlapping batch means, overlapping batch means, and standardized time series, all of which are commonly used in discrete-event simulation. For a specific CIE, the performance measures of interest include the coverage probability, and the expected value and variance of the half-length. We use both empirical and analytical methods to make detailed comparisons regarding the behavior of the CIE's for a variety of stochastic processes. All of the CIE's under study are asymptotically valid; however, they are usually invalid for small sample sizes. We find that for small samples, the bias of the variance parameter estimator figures significantly in CIE coverage performance-the less bias the better. A Secondary role is played by the Marginal distribution of the stationary process. Not all CIE's are equal - some require fewer observations before manifesting the properties for CIE validity.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1991
- Accession Number
- ADA238744
Entities
People
- David Goldsman
- Keebom Kang
- Robert G. Sargent
Organizations
- Naval Postgraduate School