Multivariate analysis.To adjust for the correlations between independent
variables, we conducted a series of multivariate logistic regression
analyses. In these analyses, the causes of divorce in Table 4 served as dependent
variables, and the independent variables entered the equation in a
stepwise fashion using a .10 probability value for the entry criterion. Because
the correlations between independent variables generally were low,
the results of the multivariate analyses did not differ substantially from the
results presented in Table 4. The only exceptions involved the associations
between abuse and having children in the household, and between substance
use and the respondent’s gender, which were no longer significant
in the multivariate models (p = .18 and .19, respectively). All other significant
results were replicated. In addition, the association between gender
and infidelity, which was marginally significant, became fully significant
in the multivariate model (p < .05).