The follow-up rate in the SDPS was over 90%, and maximum likelihood procedures were used to address missing data (Collins, et al., 2001) with Little’s MCAR test (Little, 1988) showing data missing completely at random (p = .649). Tables 1 and 3, respectively, describe AUD proband and AUD offspring demography, personality, and substance-related variables for all relevant participants combined and then separately for subjects who rated themselves as falling into categories 1–3 regarding their drinking pattern overall (i.e., deniers [Group 1]) versus those who rated themselves as categories 4–6 (non-deniers [non-deniers]). The deniers were reporting categories that might indicate to clinicians that a patient does not have problems with alcohol. The first step, univariate comparisons of Groups 1 versus 2, used F-tests for continuous variables and x2 for categorical data. Tables 2 and 4 present our key results involving backwards 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
