Among the full battery of SSAGA items, participants were verbally asked by a trained interviewer about their past instances of alcohol-related aggression, drug-related aggression, and non-substance-related aggression. After reviewing the entire SSAGA battery, we sought to include items that (A) met the definition of aggressive behavior and (B) could be could be categorized into our three aggression-related, drug-related, and non-substance-related phenotypes, based either on the content of the question or the way in which responses to the question were coded. We adopted an inclusive approach to SSAGA item selection, in which items that described behaviors that could be aggressive (e.g., throwing objects), but were not certainly so (e.g., the target of the thrown objects could have been a person or an inanimate object), were included. This inclusive approach was motivated by the relatively small number of SSAGA items that asked about aggression and the ability of our statistical models to identify items that exhibited poor empirical properties (e.g., weak factor loadings), which we could subsequently exclude from our analyses.
