In model 2, measures of parental conflict and the relationship with mother and father are added. In Model 3, the role model variables
are added. In the fourth model we include the control
variables age (only for delinquency), educational track and
financial situation. This stepwise approach allows to see
how the effects of family structure are mediated by family
processes, the role modelling and the control variables.
We limit the alcohol regression models to adolescents between
11 and 15 years old because we only find significant
bivariate associations between family type and alcohol use
for adolescents below to legal age to drink alcohol.
We present the results for boys and girls separately, but the
statistical significance of the gender differences are tested
in an overall model that includes the full set of variables.