Detecting the Determinism of EEG Time Series Using a Nonlinear Forecasting Method

The determinism of time series is investigated using a nonlinear non-parametric forecasting method. The goodness of prediction was estimated in terms of the prediction error of the predicted time series. A new definition of the prediction effect was made in the present study. Three typical kinds of time series were detected using our new method. In deterministic chaotic time series, good prediction was obtained in the new definition. However, for Gaussian random noise and schizophrenia EEG signal, the predictability could not found. We concluded that EEGs in schizophrenic patients were not deterministic