were within-subject variables. Timepoints were anchored to the time when target BrAC was first achieved. For instance, the +10 timepoint occurred 10 minutes after target BrAC was reached, approximately 30 minutes after IV ethanol infusion began. Given that this nonparametric approach is not amenable to continuous variables, we divided impulsivity scores into quartiles and created a 4-level ordinal variable. We added additional variables (i.e., test day [1, 2 or 3], gender and family history) and three-way interactions with time and dose condition involving these variables to models subsequently. Additional effects that were statistically significant at p<.05 and improved model fit significantly based on likelihood ratio testing were retained in the final models.
