A time series is a measured "output" signal, without any known, measured input. A multivariate time series is a vector-valued signal.

The analysis problem for time series and signals is to investigate their frequency contents (spectra), and/or to build predictors for them. A parametric model of a time series describes how it could be generated from a white noise source.

All system identification techniques within the Toolbox and ident can be applied to scalar and multivariate time series, simply by treating the input signal as empty.

When you import time series data, leave the box for input empty. All essential dialogs will automatically be adjusted when the working data is a time series. The spectra of the time series signal(s) are studied in the Model view Noise Spectrum. Comparisons between measured and predicted time series values are performed under the Model Output view.

Help topics.

(file iduits.htm)