We identified and combined voxels responding to both the AUDIT score and SSRT in linear regressions into a single ROI for mediation analysis. The working hypotheses were whether drinking problems lead to altered regional activations, which in turn lead to prolonged SSRT, or whether poor response inhibition alters regional activations that lead to drinking problems. Four models were built to test these hypotheses. In the first model, the AUDIT score served as the independent variable X, SSRT as the dependent variable Y, and the contrast value of voxel activity of the ROI as the mediating variable M. In the second model, SSRT served as X, the AUDIT score as Y, and voxel activity as M. In the third model, voxel activity served as X, the AUDIT score as Y, and SSRT as M. In the last model, voxel activity served as X, SSRT as Y, and the AUDIT score as M. We did not consider the other two models in which voxel activity served as the dependent variable because, as a neural phenotype, voxel activity causes behavioral manifestations but not the other way around. However, these two models were examined for completeness.
