Simultaneous Confidence Bands for Autoregressive Spectra.
Abstract
Autogressive spectral estimation has become an important method of spectral estimation in recent years despite (1) a continuing discussion of the problems of order determination and (2) a lack of easily applied procedures for determining confidence intervals or bands on the function being estimated. In this paper, we derive asymptotic 100 (1-alpha)% simultaneous confidence bands for an autoregressive spectral density assuming one has data from a finite order autoregressive process with known order or conditional upon having correctly determined the order if it is unknown. The bands are derived in section 2 and their implementation described on simulated data in section 3.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1982
- Accession Number
- ADA117072
Entities
People
- H. Joseph Newton
- Marcello Pagano
Organizations
- Texas A&M University