A total of 80 seconds of artifact-free EEG data (< ±100 μV) were selected from each recording and spectral analysis was used to extract mean spectral power and mean frequency (defined by obtaining the frequency of each .25 Hz segment within a band and dividing by the number of segments) in six bands: delta (0–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), slow-beta (12–20 Hz), fast-beta (20–35 Hz), and gamma (35–45 Hz). The spectral power data (μV2) were subjected to a log10 transformation [38], with analysis of variance applied using a mixed model repeated measures design. Between-subject factors were binge group membership with 3 levels (non-, low-, high-binging) and the 2 sexes (male, female). Midline electrodes was a repeated measures factor with 3 levels (Fz, Cz, Pz). Greenhouse-Geisser corrections were applied to the repeated measures factor with 3 or more levels, with the uncorrected df reported. Homogeneity of variance tests were conducted for the between-subjects factors.
