02 to 1.61. Table 2. Intercorrelations Between All Study Variables Overall Structural Equation Modeling Prior to modeling, we randomly assigned and averaged items for constructs consisting of five or more items into two to three manifest indicators (Little, under Cunningham, Shahar, & Widaman, 2002). For example, everyday discrimination items were randomly parceled into three indicators of the latent factor ��everyday discrimination.�� Next, we conducted structural equation modeling with latent variables to test our hypothesized model (see Figure 1). First, we estimated the measurement model for the latent variables to ensure that the psychometric properties of the measures were adequate and loaded on the hypothesized factors (Anderson & Gerbing, 1988).

Hereby, we estimated a measurement model for each construct separately, then for each pair of constructs, combining them two by two before estimating the measurement model for all the constructs in one model (J?reskog, 1993). The only exception to this procedure was with past-30-day smoking, age, SES, friend smoking, and adult smoking. We evaluated overall fit with the comparative fit index (CFI), the chi-square test of model fit (��2), and the root mean square error of approximation (RMSEA) (Hu & Bentler, 1998). The overall measurement model produced excellent fit indices (CFI = .972; RMSEA = .027, 90% CI [.025, .030]; ��2 = 702, df = 289, p < .001). After testing the measurement model, we estimated the structural model (Figure 1) (Anderson & Gerbing, 1988). The structural model provided a good fit to the data (CFI = .

924; RMSEA = .027, 90% CI [.025, .030]; ��2 = 872, df = 418, p < .001). As shown in Figure 2, standardized path coefficients suggested that acculturation was associated with higher familismo (�� = .15) and lower traditional gender roles (�� = ?.27). Enculturation was also linked with higher familismo (�� = .15), higher respeto (�� = .14), and lower traditional gender roles (�� = ?.08). Familismo and respeto related with lower levels of discrimination (�� = ?.13 and �� = ?.08, respectively), lower family conflict (�� = ?.16 and �� = ?.32, respectively), and higher family cohesion (�� = .25 and �� = .29, respectively). Fatalismo was associated with more frequent discrimination (�� = .08), more family conflict (�� = .22), and lower levels of family cohesion (�� = ?.16). Everyday discrimination (�� = .

17) Batimastat was the only significant predictor of past-30-day smoking. Figure 2. Initial structural model. Standardized path coefficients for the overall sample (N = 1,436). Dashed lines indicate nonsignificant paths. Error variances and their respective covariances are missing from the diagram to avoid complexity. Covariates were … Multigroup Structural Equation Modeling: Gender as a Moderator We examined gender as a moderator with multigroup structural equation modeling.