AUTOMATIC TIME SERIES IDENTIFICATION SPECTRAL ANALYSIS with MATLAB TOOLBOX ARMASA

摘要

ARMASA provides a new automatic spectral estimator for random data. For stationary stochastic observations, time series identification gives a better accuracy in spectral estimation than what can be obtained by FFT analysis with windowed and tapered periodograms. The parameters of the time series model accurately represent the spectral density and the covariance function of the data. The increased computational speed gives the possibility to compute hundreds of models and to select only one. The three linear time series model types are: autoregressive (AR), moving average (MA) and the combined ARMA models. The ARMAsel algorithm computes models of the three types for a large number of candidate model orders. The computer first selects the best order for each of the model types separately. Then, a single type is selected from those three models by looking for the smallest prediction error. That selected model includes precisely the statistically significant details that are present in the data, and no more.