The Theory of Standardized Time Series.
Abstract
Building on Schruben's work, this document has developed a general framework for the analysis of standardized time series. Under mild assumptions on the output process, the method of standardized time series produces asymptotically valid confidence intervals for steady-state parameters. However, these intervals are asymptotically larger and more variable then those steady-state intervals obtained by a method which consistently estimates the appropriate steady-state variance constant (such as the regenerative method). In this sense, standardized time series confidence intervals are asymptotically less desirable then those constructed by consistent estimation. These results do not, however, preclude the possibility that standardized times series may be superior in certain small sample context; this remains an area for future work. Additional keywords: Weak convergence; Stochastic processes.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1985
- Accession Number
- ADA158383
Entities
People
- D. L. Iglehart
- P. W. Glynn
Organizations
- Stanford University