Initially, we evaluated 4 different models: (a) a 3-factor model for the U.S. sample, •₍₄₀₂₎² = 918.4, RMSEA (root-mean-square error of approximation) = .08, CFI (comparative fit index) = .87, IFI (incremental fit index) =.87, and GFI (goodness of fit index) = .78; (b) a 3-factor model for the Spanish sample, •₍₄₀₂₎² = 582.2, RMSEA = .043, CFI = .91, IFI = .91, and GFI = .86; (c) a 5-factor model for the USA sample, •₍₃₉₅₎² = 878.6, RMSEA = .08, CFI = .88, IFI =.88, and GFI =.79; and (d) a 5-factor model for the Spanish sample, •₍₃₉₅₎² = 526.0, RMSEA = .04, CFI = .94, IFI = .94, and GFI = .87. Two conclusions were drawn from these results. First, the model-fit indexes were comparable for the U.S. and Spanish samples, suggesting that the data could be combined in a single analysis. Second, overall, the fit indexes were better for the 5-factor model than for the 3-factor model, suggesting that the 5- factor structure was the preferred latent structure of the inventory. Because results were comparable for the U.S. and Spanish samples, we combined them in a single 5-factor multi-group model. We fitted four different modifications of this model: (a) a model equating the correlations between the latent variables in both samples, •₍₈₀₀₎² = 1421.2, RMSEA = .06, CFI = .90, IFI = .90, and GFI = .87; (b) a model equating the correlations between the latent variables and measurement errors in both samples, •₍₈₃₀₎² = 1643.6, RMSEA = .07, CFI = .87, IFI = .87, and GFI = .83; (c) a model equating the correlations between latent variables and factor loadings of the measured variables on the latent variables, •₍₈₃₀₎² = 1627.6, RMSEA = .07, CFI = .87, IFI = .87, and GFI = .83; and (d) a model equating the correlations between the latent variables, measurement errors, and factors loading of the observed variables on the latent variables, •₍₈₆₀₎² = 1857.5, RMSEA = .08, CFI = .84, IFI = .84, and GFI = .80. According to these indexes, Model (a) described the data the best, suggesting that the underlying latent structure of the inventory is invariant across the U.S. and Spanish samples, but the measurement errors and factor loadings differ in the two samples.

The variables in the five-factor model were related. Table 2 shows the intercorrelations of the latent variables. As can be seen in the table, these latent-variable correlations are extremely high, suggesting that the five factors of the model are highly correlated and may represent a general factor, although not necessarily psychometric g, given that in past research subscales also have been highly correlated with each other but not with psychometric g.