about the cluster approach that we used in the present study. Third, the sample used here is much larger than the 1,393 subjects in Kranzler et al. and it yielded larger and more clearly differentiated clusters. Kranzler et al. found 6 clusters, each of which contained 42–350 subjects. The first three were heavy cocaine user groups, but all six clusters (including the low-cocaine-use group) contained subjects with a diagnosis of CD. In contrast, the present analysis yielded a non-cocaine-user group in which no one had a diagnosis of CD. Subjects in Cluster 1 of Kranzler et al. (N = 336) were included in our Cluster 4 (N = 3,258 subjects), which was the largest user group. Subjects in Cluster 2 (N = 303) of Kranzler et al. were included in Cluster 5 in the current study, which contained 1,916 subjects. Cluster 3 in Kranzler et al. was split in to Clusters 3,4, and 5 in the current study. Thus, we believe that the larger sample and more sophisticated methodology in the present study yielded more homogeneous clusters, which are more useful for association analysis.
