Next, we estimated a structural equation model (SEM) simultaneously regressing two latent factors (INT, EXT) on (1) 9/11 exposure (one four-level observed variable), (2) childhood maltreatment (one latent variable with five latent indicators), and (3) parental histories (alcohol, drugs, depression, antisocial behavior; four binary observed variables). All models were adjusted for demographic variables, and latent variables were allowed to correlate with each other. For each of the latent variables, the paths to the first indicator were set to 1 to establish the metric of the latent variable as has been suggested by Muthén and Muthèn (2000). As mentioned above, preliminary association analyses between 9/11 exposure and each individual disorder were conducted to ensure that single disorders were not entirely driving the effects observed in INT and EXT. Additionally, there are both theoretical and empirical (Henriksen et al., 2010) reasons to expect a particularly strong relationship between trauma exposure (9/11 exposure, childhood maltreatment) and certain disorders, such as PTSD. Thus, in further consideration of any individual disorder effects that are not captured by the latent INT and EXT dimensions, we also
