In all genetic models, consumption/use was set to missing for a given substance if the individual reported 3 or more dependence symptoms for that substance (3 symptoms being the cut-off for a dependence diagnosis). This was the case for 17.7% of female drinkers and 34.4% of male drinkers and for 8.1% of female and 12.9% of male cannabis users, respectively. By using this technique (see Grant et al., in press and Heath et al., 2002) we were able to examine the genetic correlation between heaviness of use and dependence symptomatology with the effects of dependence on consumption/use removed. (Since dependence itself leads to increasing use, an artificial inflation of genetic correlation would occur if such an adjustment were not made.) All four measures were tested for assumptions of underlying multivariate normality. Fit statistics for the alcohol measures are available in Grant et al., 2009. For cannabis-related measures, with the exception of MZ twins for cannabis use, all other groups showed evidence for multivariate normality.
