The impact of the MEG response on dynamical (\(\bf P\), top row) and topographical (\(\bf K\), bottom row) model parameters. (\(\bf A\)) For a given set of \(\bf P\) parameters, we simulated 1000 waveforms such that in each iteration the values of the original elements of \(\bf K\) (see Figure \ref{fig:MEG_wave}) were multiplied with random numbers between 0 and 10 taken from a uniform distribution. The resulting waveforms are very similar in shape and differ only in their peak magnitudes. (\(\bf B\)) Hence, the spectral analysis of these waveforms by means of FFT reveals a very narrow distribution around a single frequency of about 5 Hz. (\(\bf C\)) The time constants which describe the attenuation of the waveforms shown in (A) are also narrowly distributed (around 45 ms). (\(\bf D\)) As a consequence, plotting FFT frequency versus attenuation constant results in a very focal distribution. (\(\bf E\)) For a given set of \(\bf K\) parameters, we ran a larger number of simulations than in (A), in total 2000, because these simulations produced unstable solutions which had to be excluded from further analysis (see flowchart in Figure \ref{fig:NMflowchart}). Here, in each iteration, the stochasticity determining the elements of \(\widetilde{W}_{\tiny\mbox{ee}}\) was varied, and the elements of the diagonal matrices \(\widetilde{W}_{\tiny\mbox{ie}}\), \(\widetilde{W}_{\tiny\mbox{ei}}\) and \(\widetilde{W}_{\tiny\mbox{ii}}\) were multiplied with random numbers generated from a uniform distribution ranging from half to twice the original value. The morphology of the resulting waveforms varies hugely, which entails a much broader distribution of FFT frequencies (\(\bf F\)) and attenuation constants (\(\bf G\)). (\(\bf F\)) Hence, plotting FFT frequency versus attenuation constant shows a broad uniform distribution covering the range between 0 and 40 Hz and 0 and 80 ms, respectively.