Modeling alcohol-related and drug-related aggression items as loading onto a substance-related factor and non-substance-related aggression items as loading into a second factor yielded improved model fit over the single aggression factor model, but not past an acceptable threshold (Figure 2; Table 4). All items exhibited statistically-significant loadings onto their respective factors and both factors were positively correlated (Figure 2). Within the substance-related aggression factor, alcohol-related aggression items exhibited stronger loadings than drug-related aggression.
