The Amplitude Density Function and High-Resolution Frequency Analysis of Time Series.
Abstract
This paper describes the amplitude density function which is used to screen, identify and resolve frequencies in a time series. The technique is derived from the spectral representation of the sum of an harmonic regression function and a stochastic error process, and on the inversion theorem associated with that representation. The amplitude density function possesses the desirable properties of statistical consistency and high frequency resolution, and is related to the Fourier transform of a finite-length time series. Following the identification by the amplitude density function of the dominant frequencies in a time series, regression methods are used to fit a hidden periodicities model to the data. Also introduced is in a high-resolution frequency trace to investigate the extent and direction of nonstationarity in a time series. These techniques are illustrated with a detailed frequency analysis of the annual sunspot numbers series from 1749 to 1979, and includes analysis of a physically motivated data transformation of the sunspot series.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 06, 1983
- Accession Number
- ADA132466
Entities
People
- Alan Julian Izenman
Organizations
- Stanford University