After establishing that each independent variable (i.e., impulsive sensation seeking and parental history of alcohol problems) was related to the outcome variables, the two independent variables must be related in order to establish mediation. To examine this possibility, the relationship of impulsive sensation seeking and parental history of alcohol problems was evaluated using univariate GLM analysis. Finally, the possibility of mediation was tested by examining the relationship of parental history of alcohol problems and alcohol use variables and the concomitant alcohol-tobacco use variable while controlling for impulsive sensation seeking levels. Logistic or multinomial regressions were used for these analyses, as appropriate. Significance level was set at p < .01 for all analyses. A more conservative level (than p < .05) was chosen because of the large sample available for analyses and the desire to avoid type I error.
