as two distinct outcomes. Use of these relatively low thresholds provided greater statistical power in our multivariate models of racial disparities in alcohol problems. In our bivariate analyses, however, we utilized a third, more stringent measure to capture DSM-IV alcohol dependence, defined as having at least one symptom in at least three of seven DSM-IV domains (for further details, see Greenfield et al. (2006) and Caetano and Tam (1995)).
