EEG simulation scripts are available here, as described in: Bridwell, D.A., Rachakonda, S., Silva, R.F., Pearlson, G.D., Calhoun, V.D. (2016) Spatiospectral decomposition of multi-subject EEG: evaluating blind source separation algorithms on real and realistic simulated data. Brain Topography .
A wavelet-based approach was implemented to generate simulated EEG data. This approach is based upon the notion that continuous EEG may be decomposed as a convolution of a series of basis functions (i.e. wavelets) which have defined temporal and frequency properties. The distribution of the associated coefficients was estimated within select frequency bands from real data. Then, simulated wavelet coefficients were generated by randomly drawing samples from that distribution. The simulated coefficients were reconstructed within the separate frequency bands, generating simulated EEG data with temporal and spectral properties that are consistent with the EEG segment that was used to estimate the coefficient distributions.
Within each simulated subject, wavelet coefficients were randomly drawn from a logistic distribution derived from wavelet coefficients estimated from real EEG. Thus, we assume that the spectral characteristics are common and the time courses are uncorrelated across subjects, as expected for data collected in the absence of an explicit task (i.e. during rest).
Simulated scalp EEG topographies were generated by assigning electrode location(s) as electrical current sources and sinks, and interpolating across the neighboring electrodes. The simulated time series is assumed to represent the time series of the electrical current source, and multiplying this time series by -1 generates the time series of the electrical current sink.