The differences in self-reported frequency of DUI behaviors
between males and females led us to introduce gender (dummy
coding: males = 0,females = 1)in block 1;theAISS subscales in block
2 and the process variables of the Fishbein–Ajzen model (viz., DUI
attitudes, subjective norms and perceived behavioral control) in
block 3 for better control of their effects. Alcohol use was included
inthe last block. The input sequence for these variables was selected
consistently withourprimary aim(viz.,to assess the contributionof
process variables to sensation seeking in predicting drunk driving
by youths).
As can be seen from Table 3, the model explained 53% of the
variance in DUI behaviors. The first block (R2 = .08) revealed a significant
contribution of gender to explaining drunk driving. The
Sensation Seeking scale of the AISS, which was included in block 2,
explained 6%ofthe variance. Self-efficacy in avoiding DUI behaviors
and perceived disapproval of such behaviors by peers contributed
substantially to explaining risky behaviors (R2 = .36). In any case,
including these variables in the model reduced the contribution
of demographic and personality variables to explaining DUI behaviors.
Finally, including drunk driving increased the predictive ability
of the model by 3%.
The potential mediational relationships between sensation
seeking, the variables of the theory of planned behavior and their
contribution to predicting drunk driving were also examined.
Mediation exists when a predictor X affects a dependent variable
Y indirectly through one or more intervening variables or
mediators M. The effect of X on the proposed mediator is called
path a, whereas the effect of M on Y partialing out the effect of X
is called path b. The indirect effect of X on Y through M can thus be
quantified as the product of a and b (i.e., a·b). The direct effect of X
on Y after controlling for M is c
. The simple relationship between X
and Y is often referred to as the total effect of X on Y and quantified