As seen in Table 3, the overall parsimonious regression model was significant, χ2(11) = 210.21, p<.001. Significant higher hazard rates were obtained for six of the ten predictor variables including race, first alcohol dependence symptom, and first conduct disorder symptom. The presence of a biological alcohol dependent relative (aunt, uncle, or grandparent), an alcohol dependent biological father only, mother only, or pair of affected parents, educational background of mother, and mother's age at time of participant's birth were also significant predictors. Note that educational background of mother and having a teenage mother became significant in the parsimonious model once the other non-predictive variables were removed from the analysis. This is likely due to the collinearity among mother's educational background, being a teenage mother, and metropolitan status. That is, the standard errors of these three terms in the model were slightly inflated due to the collinearity among the variables. Removal of the one variable, metropolitan status, by reducing standard errors and increasing slope estimates, resulted in the other two variables becoming statistically significant.
