elimination logistic regression analyses using variables that significantly differentiated between deniers and non-deniers in Tables 1 and 3. Finally regarding methods, for both probands and offspring data, multicollinearity was assed using both simple correlation matrixes among the variables and evaluating for variance inflation factors (VIF). For correlations, values greater than or equal to .80 and for VIF values greater than 5 indicate possible multicollinearity (O’Brien, 2007). For probands there were no correlations greater than .80 with the largest r = .47 and the largest VIF = 1.64. For offspring, out of 171 correlations, 2 were high (smoking with age = .80 and dependence item 6 (D6) with abuse item 3 (A3) = .89), but almost all VIF’s were less than 2, except for 3 with values up to 2.47. Thus, multicollinearity was not apparent in these data sets.
