STATISTICAL SPECTRAL ANALYSIS (SINGLE CHANNEL CASE) IN 1968.

Abstract

Statistical spectral analysis is a technique for data analysis which computes from observed functions of time various functions of a variable called frequency. The record can consist of a single function X(.) of time (single channel case) or of several functions of time X sub 1(.),...,X sub n(.) (multi-channel case). This paper describes the view that to understand statistical spectral analysis in 1968 one must comprehend three distinct aspects: (1) how to define the spectrum, (2) how to compute the spectrum (four methods are distinguished: filtering, smoothed periodogram, covariance averages or filtered periodogram, autoregressive), and (3) how to interpret the spectrum (especially with regard to testing for hidden periodicities, estimation of the spectral density, and mixed spectral estimation). The effect of Fast Fourier Transform techniques on statistical spectral analysis is also discussed. A basic theorem on the means, variances, and covariances of filtered sample spectral density functions is stated. (Author)

Document Details

Document Type
Technical Report
Publication Date
Jun 10, 1968
Accession Number
AD0671885

Entities

People

  • Emanuel Parzen

Organizations

  • Stanford University

Tags

DTIC Thesaurus Topics

  • Computational Processes
  • Computing-Related Activities
  • Covariance
  • Data Analysis
  • Data Mining
  • Data Science
  • Fast Fourier Transforms
  • Filtration
  • Frequency
  • Information Science
  • Mathematics
  • Periodic Variations
  • Spectra

Readers

  • Statistical inference.
  • Theoretical Analysis.