Alpha was set at .05 for all analyses. We first obtained descriptive statistics and correlations between key measures and smoking characteristics. We then ran linear regression models in which AS served as the predictor and one of the smoking measures served as the outcome. For each outcome, we then tested for mediation by using the product of coefficients from the A and B paths as the indicator of the strength of the indirect effect, for which significance was determined using the RMediation package (Tofighi and MacKinnon, 2011). To determine the effect of AS beyond what is accounted for by mediation (i.e., the remaining direct effect), linear regression models were run examining the relation of AS to a smoking variable after controlling for negative urgency.
