A series of repeated measures analyses (trajectory class × wave) were then conducted (using SAS PROC MIXED, SAS Institute Inc., Cary, NC) to compare developmental differences among the impulsivity classes in three separate indices of alcohol involvement. Participants who reported abstaining from alcohol across all assessments (n = 6) were excluded from all analyses. Sex was modeled as a fixed covariate in all analyses. Of particular interest were comparisons of the developmentally limited impulsivity class (Class 3) to the other impulsivity classes (especially the relatively high and steady impulsivity classes 4 and 5; see Fig. 1). Thus, mean changes (as well as cross-sectional differences at each wave) in alcohol involvement and related outcomes were compared among impulsivity trajectory groups at each successive wave (i.e., between ages 18 to 25, 25 to 29, and 29 to 35). To reduce skew, all alcohol variables were square-root-transformed prior to analysis (Tabachnick and Fidell, 2001).
