Each model was re-tested adjusting for covariates. Because AS and anxiety symptoms are closely related and have both been tied to negative reinforcement-related smoking motives (Gregor et al., 2007), MASQ-AA was included as a planned covariate in order to examine the predictive validity of AS on smoking behavior beyond that of its shared variance with anxiety symptoms. We also included FTND as a planned covariate to determine whether relations between AS were unique from heavy and compulsive smoking and qualitatively specific to other smoking variables. Results are reported as standardized regression coefficients (βs). Effect sizes for mediational analyses were expressed as percent of the total effect, which is the mediated (indirect) effect divided by the total effect (i.e., the indirect effect divided by the sum of the indirect and direct effects; MacKinnon, 2008, pp. 79–82).
