We computed power spectral densities up to 400 Hz to see if we could replicate earlier findings of beta band reduction with dopaminergic medication and whether we could observe HFO in the recordings. Furthermore, we used the power spectra to determine if PAC could explain any variance in clinical scores independently from spectral power alone. We used Matlab’s (R2014a, The Mathworks Inc., Natick, USA) pwelch.m function to compute power spectral densities for each epoch (using default values of 8 Hamming windows with 50% overlap), and then averaged spectra across epochs. Values for frequencies ±3 Hz around 50 Hz harmonics were removed from the spectra in order to avoid the influence of notch filters on summary measures of spectral power. Power values for frequencies below 50 Hz were normalised by dividing by the power summed over all frequencies in the entire spectrum for each subject, condition, and bipolar STN channel individually. Values within the 150–400 Hz interval were divided by the mean power between 400 and 500 Hz to eliminate inter-subject variability in offset values in this frequency range.
