We tested max drinks and DSM-IV symptom count phenotypes for genetic association. These data were analyzed using linear regression models implemented in SAS v.9.1 and PLINK (Purcell et al., 2007). In all models, genotypic effects were modeled additively as the number of minor alleles increased. Due to its skewed distribution, max drinks were log transformed to better approximate a normal distribution and analyzed in a linear regression model. The number of DSM-IV alcohol dependence criteria endorsed was also analyzed using a linear regression model. To better approximate a normal distribution, the maximum symptom count was set at five, and 60 participants who endorsed six or seven symptoms had their counts truncated at five. In all models, we adjusted for the effects of gender and nicotine dependence, and also for age in the max drinks analysis. Covariates were included in the models if the associated P-values were ≤ 0.05. Eight participants who reported never having consumed an alcoholic beverage were excluded from analysis.
