The results presented here do not generalize to the overall population, since non-drinkers were excluded from our core analyses. By restricting the analyses thus, we adjust for the disproportionately high rates of abstinence among African Americans and Hispanics. Were we to include abstainers, this would likely result in the artificial deflation of disparities at the no/low level of heavy drinking, since this is where abstainers would be grouped; we would expect little impact on the multivariate results for moderate and high levels of heavy drinking.
