After evaluating the initial mediational model, we examined whether any structural path varied as a function of gender, race/ethnicity (Caucasian, Hispanic/Latino, or African American), and adolescents’ living arrangement, using a multiple group model approach (Muthén & Muthén, 2006). This approach entailed two steps, the first of which involved estimating an unrestricted model where all structural paths were free to vary across gender, race/ethnicity, or living arrangement. The second step involved estimating a model in which we constrained the structural paths among variables to be equal across gender, race/ethnicity, and living arrangement. We then used a χ2 difference test (Δχ2), as described by Muthén and Muthén (2006), for use with WLSMV, to determine whether constraining paths across gender, race/ethnicity, or living arrangement resulted in a significant decrement in the model χ2. If constraining the paths to be equal across these variables resulted in a significant decrement to the model χ2, we could then assume that the relationships among them differed.
