The present application of multigroup IRT modeling to diversity of substances used allowed us to examine both the structure of poly-substance involvement as well as how age and sex impact poly-substance use at the level of the underlying dimension. Model fit indices supported a single dimension of poly-substance involvement during adolescence. This extends previous findings regarding adult poly-substance involvement (Kirisci et al., 2002) to the age range when substance use develops. Moreover, the clustering of substance locations in a relatively narrow range of trait levels (see Figure 1) indicates that diversity of substances used generally provides information about involvement with multiple substances. The largely non-significant differences in substance discriminations (see Table 3) demonstrated that all substances provided similar amounts of information on the underlying dimension (with the exception of uncontrolled substances; i.e., over-the-counter medications and inhalants; which were particularly uninformative).
