Fitting Models Of Nonstationary Time Series: An Application To Eeg Data