The study hypotheses were tested with a series of t-tests for each metabolite in each voxel. Initial tests compared the control group (N=15) vs. the total AUD group (N=20), and follow-up tests compared the control group vs. each of the three combinations of AUD subgroups: the 10 AUD-R, 15 AUD-R+AUD-lost, and 5 AUD-A. The effects of categorical variables on metabolite levels were examined with t-tests and simple regressions; relations between continuous variables and metabolite levels were tested with Pearson correlations (r) and verified with nonparametric Spearman Rank-Order tests (ρ). Family-wise Bonferroni correction for 4 metabolites with directional hypotheses required p≤.025 with α=.05 for statistical significance.
