Informace o publikaci
Spectral Density Estimation via Autoregressive Modeling
Autoři | |
---|---|
Rok publikování | 2014 |
Druh | Článek ve sborníku |
Konference | Mathematical Models and Financial Mathematics - Book of short papers |
Fakulta / Pracoviště MU | |
Citace | |
Obor | Aplikovaná statistika, operační výzkum |
Klíčová slova | ARMA models; identification; order selection; parametric spektrum; spectral estimation; time series |
Popis | The spectral density function is a commonly used tool when analyzing time series in the frequency domain. Parametric spectral estimation methods have gained attention as potentially interesting tools in the last four decades. They allow the improvement of the statistical properties of spectral estimators with respect to the Fourier-based methods. Estimation of the parameters of ARMA and MA models needs the resolution of a set of nonlinear equations, whereas the AR parameters estimates can be calculated by solving a set of linear ones. Moreover, algorithms, such as Levinson's, used to solve this set of equations are computationally efficient. When the AR modeling assumption is valid, spectral estimates are less biased and have lower variability than the Fourier-based ones. For these reasons, the AR method became the most popular approach to parametric spectral estimation. |