The present sample of Minnesota twins is not as diverse as the US population, and so these results may not be fully generalizable (e.g., the prevalence rates and relationships among substances may differ in other populations, especially where access to certain substances differs). Future research may address this issue through replication in a large national survey. In addition, to ensure that our data conformed to IRT model assumptions, we dichotomized use of each substance and clustered less frequently endorsed substances (i.e., other controlled substances and uncontrolled substances). However, this was a longitudinal study, which allows for greater interpretability of developmental effects. With these considerations in mind, we turn now to describe our findings in terms of the structure of poly-substance involvement, how development and sex affect that dimension, and the strength with which individual substances relate to the dimension of poly-substance involvement.
