Independent samples t-tests across the skeletonized FA map permitted voxel-wise comparison of mean differences between binge drinkers in controls (Cox, 1996). To control for type I error, a combined t statistic magnitude (p < 0.01) and cluster volume thresholding method was utilized. Monte Carlo simulation revealed that 27 contiguous voxels exceeding the t-threshold were required to protect family-wise error at p = 0.01, assuming 4-mm FWHM intrinsic smoothing (Smith et al., 2006). No additional smoothing kernel was applied to the data. Areas representing group differences of FA on white matter tracts were identified using a DTI atlas (Wakana et al., 2004). To further explore the influence of alcohol use on white matter integrity, follow-up bivariate correlation analyses (using Pearson's r) related binge drinkers' alcohol involvement variables (lifetime hangover symptoms, estimated peak blood alcohol concentration, lifetime drinks, and duration of drinking) to mean FA in clusters that significantly differed between groups.
