largest possible mutually exclusive samples of individuals with DSM-IV and DSM-5 diagnoses. Using abuse/moderate AUD as an example, the sample for DSM-IV abuse comprised all respondents who were positive solely for abuse (group 1 above) and half of those positive for both abuse and moderate AUD (group 3), the latter upweighted by a constant adjustment factor of ≈2 (the inverse of the split group 3 sample size divided by the full group 3 sample size) to be representative of its full unsplit prevalence. The sample for DSM-5 moderate AUD consisted of all respondents who were positive solely for moderate AUD (group 2) and the remaining half of those positive for both abuse and moderate AUD (group 3), the latter again upweighted to its full prevalence.(Without upweighting, the profiles of DSM-IV and DSM-5 AUD would have overrepresented individuals positive solely for the disorder in question, all of whom contributed to the profile compared with only half of those positive for both disorders.) The same approach was used to create mutually exclusive samples for DSM-IV dependence and DSM-5 severe AUD. In order to create the two random half samples required for this approach, we applied even case identification numbers towards the DSM-5
